Speed adjustment considerations

[low_grade]

Just thought of a comparison test, implemented with a load of Coal Miner's Daughter fresh on 1885, no receipts (put the game on very slow in December 1884 to get that.) Replace all locos with new Americans, run for a year, record average speed (15 mph = 180 miles on average.) Then reload and replace all with new Dukes, then Connies, then Stirlings, now recording also loads delivered, express and freight revenue, and train maintenance and fuel costs. Then redid the American to get the rest of that data (and discovered results could vary considerably due to random events I suppose, 10% different results the 2nd try!) Then I thought of a whole bunch of ways in which my data was useless unless I isolated a lot more variables, and got frustrated trying to get access to all locos to test also the later Camelbacks etc in this same setup, and gave up.

But one really interesting thing to note... this is a map with a LOT of tough grades, and I had been running all Americans, so it's not the usual 8 or 7+cab or 6+cab+diner, uphills were largely 4+diner up and 6+diner down, long distance with only some moderate grades 6+diner, short distance 5+diner and just 4+diner between e.g. Hidden Temple and Black Diamond Mine. I used Gumboot's revised dining cars and without recalling the specifics I believe it is lighter and so just threw it in. Keep this in mind, as I didn't change consists, and well Connies probably would have all been 7+diner, Duke and Stirling not sure, probably Duke about the same as the American and the Stirling a bit more conservative on the uphills, probably just 3 no diner.

The result that surprised me: the Stirling was the best, on this map! Oh, and tested the Vittorio Emmanuel, so out of 5 locos, best profit ($16.5M on 147 locos with the others at ~$14.5M with the Duke much lower at $11.7M due to very high maintenance and slightly higher fuel costs and maybe bad luck that year,) best speed (19mph vs 17 for the Vittorio, 16s and 15s for the rest) and most loads hauled (1763 vs 1650 for the Vittorio, 1600-1485 for the rest.) An extra $2M+ on say 150 locos would justify spending only $12k more though, 15k tops. So I guess the Stirling isn't recommendable as a wholesale replacement, even though it does better on average.

Also, it was a doomed comparo to begin with since the test map I had handy also eliminated breakdowns and crashes, so reliability didn't come into play. But still, it's just in my mind so predictable as to be pointless even testing except to see how bad a loco could be on this map full of big 6 grade climbs to see how the Stirling would fare, of course it would be reduced to a crawl and tangle up traffic! So how the hell did it soundly beat the others?? What am I not understanding to have so grossly underestimated its chances? This is otherwise a very diverse map, with about as many short and long runs, flat, moderate, and steep runs, some express and many dedicated haulers but mostly autoconsist.

At the very least I have to be convinced that the Stirling has much broader applications than I had though.

[RulerofRails]

I don't really trust the game's average speed reading, the same way I don't trust the load miles hauled figure. But it's decent as a ballpark. I have come to similar conclusions that on an AVERAGE hill-climbing isn't a very important consideration when using one loco in an all-purpose role (for everything). Even on the "mountainous" maps, most locos wont really face steep grades if you route track cleverly. Hill-climbers are a niche. On a handful of specific routes they really shine, so the best use (on auto consist) would be to have a dozen or so for those routes and the other 50-100 locos whatever offers the best complete package (running cost/performance).

Part of the perception for poor performance of the Stirling (and other engines with relatively low pulling power) is the skewed figures the game provides in the performance readout as a result of using the "future" heavier car weights.

Purely for testing purposes, to easily get any train you want just use the cheat "trains are in my blood". Warning with that: you get all locos, you can't get rid of any from the list (for ex. all the diesels or the electrics). Which is good because then people who really cheat this way can't hide it in a saved game , but it means you have a long list. But no problem if doing a mass-replacement.

PS.
Off topic (CMD), what do you think of the random test in the events governing goods and coal slow-down at the Daughter mine? Final test is underway.

[low_grade]

I know nothing of the cheats, so that's a big help for testing! Now to pick another balanced map I have in my saves that allows reliability to come into play... and probably run it through a year three times with each loco from that period and take an average, then open the editor and change the start date to another period and run those locos, eventually running all available in A, B, C, D, and E... That's maybe 300+ test runs, taking about 5 minutes each with data collection, 1500 minutes... oof this is going to take a while, lol! let's say an hour a day and I'll be done in a month.

Now, what could I learn? And before embarking any suggestions?

And reviewing my maths while sober... Faster trains hauling more cargo will speed the gradient collapse and reduce profitability, so really you'd run fewer faster locos to maintain the price gradients and profitability. So what can I see with a one year comparison? And reviewing my initial findings, what I'd rather look at is return on investment, and given gradient collapse with faster locos my goal would be 20% ROI in the first year.

ah, keep in mind there's a 10% loco speed boost in effect also skewing this map's results...

So, American (my edit I believe) 55k, 7k maint, 54mph top speed, Poor fuel economy, Below Average accel, power with 8 freight is 29mph at 0%, with 1 freight 18mph at 6% (according to the game), 55k*147=$8,085k
Vittorio (not edited) 120k*147=$17,640k
Connie (my edit again) 100k, 9k, 53mph top speed, Below Average fuel, Poor accel, power as above the game states 43mph and 32mph, 100k*147=$14,700k
Duke (possibly my edit?) 100k, 9k, 61mph top speed, Average, Average, power as above 40mph and 18mph, $14,700k
Stirling 140k, 8k, 83mph, Below Ave, Above Ave, power 27mph and 4mph (so not terrible at pulling long trains on the flats with the real stats...) 140k*147=$20,580k

And yeah I should probably test separately my edits, Lirios, base 1.05, 1.06 stats, and whatever other user created edits are available (I need to grab your edits of the Big Boy and Challenger...)

So I got an extra $2M that year over the American, and only got the stats from the year the American ran better, let's just say that the American on average would run more like the duke, and the Stirling got about $5M more than the Duke. So, $5M from an extra $12M outlay in train costs, okay yeah, that's a good investment (40% ROI!) Even $2M isn't bad, 16% ROI.

Hrmmm... but as stated before I'd probably run fewer Stirlings and keep profit stable, which with no ROI is a strong argument for just running Americans, because running more of those will still be cheaper than running fewer Stirlings for the same profit...

The truth however is a complex combination of the two arguments. Americans are essential for cash-strapped starts, but I'd have to do like a 5 year comparison to see if price gradients really flatten out/to see if the Stirling remains much more profitable after years of running. I suspect it will maintain at least a $2-$3M advantage over the American (Ultra Cool on pax heavy maps will help a lot with this) giving a 16%-24% ROI over the American on this map. The Connie wasn't too bad, and the Vittorio surprisingly good despite being Ugly, and I suspect both of them give decent ROI over the American, too, but maybe more like 10% (Connie was $6.7M more, with about $3M more profit than the American, so almost 50% ROI! but over 5 years...)

Okay, now to try on another map, I'd like the Connie's reliability to come into play...

[RulerofRails]

Another useful cheat for testing is time travel (works only in 1.06). Milo added it. The code is "wolverine@msu" Should be used with caution when going backwards.

Faster trains hauling more cargo will speed the gradient collapse and reduce profitability, so really you'd run fewer faster locos to maintain the price gradients and profitability.

Spot on. In a typical mass replacement for a faster engine we see a spike in profits. What is happening is that the price demands are more effectively equaled in the short term. Then there's a lull since there's little demand for transport, eventually a new price equilibrium is established. Also, breakdowns disappear and some engines may skip a service stop or two depending on their proximity at the moment of replacement.

My "thinking" at the moment is that a non-shattering performance edge for example a higher acceleration level, or +25% top speed can largely be played around (especially in the early years) in the typical game. It comes down to number of locos and how many stops per route.

For freight: the job of the loco is to keep prices at equilibrium, distributing cargo across the map. Up to a certain point, on an identical route setup the slower loco will encounter a stronger price differential on each trip. So profits stay pretty close. IMO, 80% of my trains aren't running at full capacity in the typical scenario. That's if I consider that I want them hauling $15k+ (just making these levels up for comparison) loads. There's the option to run heavy service with efficient trains and think that $8k+ loads are good. Only feasible in the 19th century. Partially then, but definitely in the 20th century ROI is against this method. But the main argument against this play style is that it's very likely you will steal cargo from industries just based on how far they are from the station, hobbling production at those that are otherwise reliable. Which in turn reduces the quantity of high value cargo you have to haul.

For express: it matters less. If running with Hotels those $8k+ loads are worthwhile. Speed helps express more since these cargoes are more time sensitive. I also think that typically the highest express revenues are unlocked when using efficient and the system is over-serviced. In your test (not saying you over-serviced etc.), I'm pretty sure extra Express revenue thanks to the Ultra Cool passenger rating played a part in the Stirling's success.

---

Not going to discourage any tests you might run. I will say it's really difficult to get a controlled test environment. Where I would like to go with locos is so that various ones can be used on the same map. I have done similar tests (rigged up Orient Express map), but my thinking is that half the point is to encourage diversity in loco rosters. As such a one-loco-for-all-applications approach isn't the end goal. Useful for gathering some stats (in my run the Duke and Connie were fairly evenly matched, with a slight edge to Duke, but map wasn't as hilly as this one and using the 60mph Connie breed) to describe the current situation, but I thought that it's probably not the metric I really want to use as it reinforces the one-loco-for-all-applications mindset. To allow a real choice (you wont lose your shirt if you run a Northern instead of a H10) between some locos I'm almost sure that advantages between them shouldn't be clearly defined in such a test. So maybe if there's no apparent advantage the stats are ok?

I've been meaning to get an acceleration curve generator up for awhile. Maybe I'll get time for that here in a bit.

[Gumboots]

I used Gumboot's revised dining cars and without recalling the specifics I believe it is lighter and so just threw it in.

Did you use the entire express pack (they were all packed together) or did you extract the dining car and only use that?

The reason I ask is that all of the express cars are lighter than default. They are not intended to be used with default locomotive stats.

[RulerofRails]

To put my feeling (pretty much what I said above) a different way: when I play the game I rarely spend the time to watch one specific loco for long enough that I'm satisfied that its workload is 100% correct. I often buy two at the same time, and while I do buy some for short term stuff before switching to long term routings, in either case it's very rare that I will go through my existing engines some years later to re-evaluate their workload even though surely economic potential has increased as time has passed. So I rely on observation, feeling, and/or guess-timation in the majority of cases for workloads on the bulk of my locos in a typical established game.

If I see an obvious problem such as heaps of cargo escaping (this can be counter-productive as the cause may be breakdown or other traffic so I risk buying excess locos) or a train making no money at all I will make a change, but it's generally not the focus of play. The only exception is in the rare economically tight games (typically maps with scarce resources) and in that case I will only have a few locos so the process is manageable. Gets tedious with large numbers of engines anyway.

The things that seem to make the difference in my game play are the strategic calls and the chase of ROI on big ticket items like new expansion or industry build/buy. Trying to get engine workloads 100% optimized is likely just a distraction (my mind isn't that big ) that will impede my performance on the larger scale.

So the point being that on paper a certain engine has an advantage. It can do more work. But this is limited by my ability to correctly estimate (best I could do would be to watch each engine carefully, but even that's playing catch up) workloads and adjust rosters/service patterns to compensate. It's pretty easy to lose a portion of raw advantage here. When the difference is smaller any advantage can be completely lost in the noise. Given that the majority of my recent (last 3 years) game time has been using the slower, cheaper locos (as determined by fuel cost, and also because short term ROI can be higher), my usage has no doubt evolved to get the most from them.

I fear that there is no realistic way (at least with my skills) to precisely determine the economic advantage realized in-game from a certain level of better performance. Worst case scenario is just to put fuel cost per mile at equal (or within a small range say +/- 10%) for a given time period and leave it to player skill to find the potential economic advantage.

[low_grade]

@ Gumboots I used the whole pack, I was pretty shocked when I found out how much dining cars and caboose (I kinda think this is a word, like deer, that should be able to be either singular or plural...) weighed (as much as freight cars, wasn't it?) In my mind, I'd thought I'd read or maybe just assumed they were much lighter and the decision was not about how much fuel and speed will be impacted but potential revenue lost from not hauling paying passengers in those last two slots. So yeah, I appreciated your reductions in weight, I'll have to add that to RoR's spreadsheet I've had open for the past month on my desktop where I put the numbers in and see what my real speed graph looks like, and expenses/when I should replace.

@RoR, yeah, I've tested a few more on a new map and immediately realized, no breakdowns or crashes... I'm going to have to run to 5 years and see what those numbers are, on a collapsed/equilibrium system. That's obviously not an ideal situation to consider, since you will never be in that situation in a game, you will always be developing. Well, sometimes I cruise for the last few years to meet goals, but by then success is sure and I don't need help deciding which engines to use... Meh, will probably still run some 5 year trials.

But mainly also, as you've said, how does this help me to play with more locos? Well, I'm not sure, mainly I'm just experimenting because I had an idea and who knows what I may learn. But the other thing I realized is that I need to get specialized maps that stress one type of situation or another, right now I'm doing Russia in 1876, holy crap what a flat map, a bit bumpy in places, but probably more 0's than anything else, and no long grades, only a few 3's and maybe a couple 4's on the whole map. And fairly long runs on average. My ideal would be to get maps with mostly flat, moderate, or steep (the Coal Miner's Daughter, e.g.,) and that are either express heavy or freight heavy, and that have mostly short runs or mostly long runs, and which are either resource dense or resource sparse (although this may be the same as express or freight heavy...,) and get to a point in the games where a network is mostly mature, and you're now running everything on auto.

Hmm, maybe get a save from late December instead of first thing January, replace all, that way it already has a year on it and reliability will begin to come into play, then run it for 2 years, hopefully haulage won't stagnate by then and it will be more representative of real game experience...

Also, some games I will micromanage, run on slower speeds to catch every arrival and decide what to do next, but this is usually because I've started with a bad strategy and every penny counts, lol! And I try to stick it out rather than restart. But mostly I have a very set development pattern, and most scenarios I will set almost everything on regular auto consists and never examine routing again. I will every year copy the best routes (over $300k or $200k, top 10% or so,) and a couple times a game I'll go through all my locos and just see if they make sense, and end up deleting a couple and rerouting a couple.

I fear that there is no realistic way (at least with my skills) to precisely determine the economic advantage realized in-game from a certain level of better performance.

Doesn't stop us from trying, though, does it? But I think of it as part of the game, figuring as much out as possible to be able to make the best decisions. Not perfect decisions, that's not a game, that's a logic puzzle. There's enough complexity here that at the end of the day you have to make a decision based on your gut or best guesstimate. But is that decision 80% knowledge 20% guess? Or 20% knowledge 80% guess? Yeah, I'd rather know more when decision time comes.

My "thinking" at the moment is that a non-shattering performance edge for example a higher acceleration level, or +25% top speed can largely be played around (especially in the early years) in the typical game.

This was the other thing that I was struck by in my initial tests, the variance in average speed was not what I expected, they all managed about the same. Gotta say I do trust the average speed in the Status, it matches the other statistics like loads delivered and revenue fairly proportionally. But yeah, I'll probably keep going with this. Round two revealed Crampton and Firefly are great, not surprising, but reinforcing my conviction that especially in dense traffic acceleration is the most important stat for a loco. I'll redo the save file to late December and try again with 2 year runs, that'll be a pretty good data point, and if you have any save files from any time of year from various periods and variations of the variables I defined above please attach them so I can try in the different conditions I'd like to examine.

[Gumboots]

@ Gumboots I used the whole pack, I was pretty shocked when I found out how much dining cars and caboose (I kinda think this is a word, like deer, that should be able to be either singular or plural...) weighed (as much as freight cars, wasn't it?)

Defaults are 1/3 more than freight, which is totally bonkers.

Round two revealed Crampton and Firefly are great, not surprising, but reinforcing my conviction that especially in dense traffic acceleration is the most important stat for a loco.

I don't mind the Crampton for its era, as long as the terrain is flat. I've even used it to haul freight when trying to get haulage stats up on flat terrain.

The Firefly is a bit of a tricky case, due to its exceptionally bad reliability. I find that using them in large numbers becomes too costly, since if I keep them longer than 3 years they start breaking down all over the place. My preference is to use a few on the most rewarding routes for them, and use Adlers (earlier) and Beuths (later) for the rest of my roster.

[RulerofRails]

Ok. Here's the Speed Curve Generator. This is quite handy for getting an idea of the real speed of a loco. The limitations are that it is working with acceleration from a standing start in constant conditions (corners, bridges, change of grade not represented).

Open to suggestions on scale (if you know a little about spreadsheets you can do it yourself too). Even the slowest* acceleration level, #1, Extremely Poor reaches Max Speed after only 112 slow-time hours (half- normal time days). Technically it's only a curve until Max Speed is reached. For the quicker levels that's not very long on the graph.
*I haven't actually tested the unused "Is it moving?" corresponding to level #0. Maybe I will someday. In any case this sheet doesn't support that.

Obviously we have a linear line after Max Speed is reached, so for any longer distances this just gets extrapolated (only needed for slow engines). I limited time to 200, but distance is free. Levels to remember, a steamer runs out of water at 227 miles, any train runs out of oil at 610 miles (definitely not recommended practice).

I included 6 entries. They are all graphed. I noted in the tips that putting 0 for max speed will graph that entry flat on the x axis, anybody can do that to get a less cluttered view. Not hard to add more if you know just a little about sheets, or I could add them if desired.

The sheet was created in LibreOffice but saved in XLS format as well. Formatting is at a minimum so should work well with Excel.

[CeeBee]

I looked around and couldn't find the caboose/express car pack that is referenced. Any chance one of you fine gents could point me in the right direction?

[RulerofRails]

Those are found within the large "Cargo weight revamping: express/freight differences" thread. Currently they are attached to these posts: Caboose pack: : : Express pack : : : Bug-fix for express (in add this to the pack).

Please note that these use the car weights from the new scale which are significantly lower than the default stats. They are NOT designed to be used with the default loco stats. Need to use edited loco stats. This isn't done yet. Testing is underway to try to do the best we can with the adjustment to the loco stats. Because there is no full roster of cars to use yet, there's not a big rush at the moment. Just sussing things out, which is what threads like this are all about.

[CeeBee]

Thanks for explaining the situation. Probably best for me to wait till things are finished.

[low_grade]

Wait, why not for the default loco stat? I've been using them for a month I'd guess haven't noticed any weirdness...

And slow-time hours, so 60 hours in a month? I'd been assuming 1 hour = 1 year since the few slow-time scenarios I've played have taken maybe a couple or few days to complete? And felt like maybe a 12 year win or 15 year, not too long. Hmm, never did the mental math, that would be only 12-15 hours, not a few days. But 60 hours = 60 years, not a month. By my feeling and recollection, an hour slow time would be close to 60 hours = 15 to 12 years, so one hour = 2.5 to 3 months. Also taking the games average speed measurement, and by the miles per year estimates you have 400 miles per year in the D era with trains averaging 33 miles per hour say (both reasonable estimates) then clearly an hour is a month. So I don't understand something here...

And, a mile is one of the smallest squares on the map, or a track segment? I knew Oil had less than half the demand of water, and I think sand depends on grade but is more or less the same as water (less on flats, more on grades, even downhill,) but I never knew the exact demand figures, cool! Nice graph, but if only the game would count how many times trains have been stopped by traffic in a year, then we would know what distance to use. Of course we can set up only one train per track between each town and double track to get an ideal where water towers and oil are placed perfectly with your closest-to-town model and then the slower-to-get-going train would win out... but I maintain that acceleration is king. In my games, more trains=more profit, and I saturate the rails as much as I can, and there are traffic jams all over the place. In the testing I'm doing I happen to have the save set so I'm staring at one of the major hubs on the Russia map, Voronezh, and dang trains are getting stacked up there! But the faster accelerating ones clear it out much quicker, and watching time on superfast, this is months of difference in time before some trains can escape. So given all the traffic, I'd say for me on a developed map much of a loco's time is spent on very short trips, sometimes only getting a few track segments ahead before somebody else comes up behind, and many more trips are probably only 40-50 miles, and the average trip maybe less than that given all the very short trips in congested areas. Granted the big money-making hauls just plow right through.

Also watching my tests, I'm skipping to watch different routes each time, and I'm consistently noticing, junctions near cities are bad, and water towers, too. Maybe okay with quick accel, but if you're considering running poorly accelerating trains, design your network accordingly.

And awesome time-warp cheat! So easy to test 4 year old engines!

And so, finished the B-period test on the flats in Russia, really many long runs here shouldn't punish poor acceleration too much, even with dense traffic in the major hubs, after watching many trains make long uninterrupted runs. A bit bumpier than I'd earlier described, lots of 2's, quite a few 3's, and a few 4's here and there, but mostly 0's and 1's. Most remarkable perhaps is the Camelback's fuel costs, wow! Also further validation, I love the Eight Wheeler!

[Gumboots]

Wait, why not for the default loco stat? I've been using them for a month I'd guess haven't noticed any weirdness...

The idea of them is to get a better split between freight and express, so that express locos can haul express like they actually did without them being so strong in the game that they become the default for all consists. WP&P's custom express cars suffer from this problem. Some of them are really heavy, so if you have an express loco that will do a good job with them it will also haul full freight consists like a rocket. He got too worried about historical accuracy and forgot about game balance.

Due to the game's limitations we can't make freight too heavy (runs into problems with reliability, etc) so making freights slightly heavier over time, and express lighter, is the best way to tackle it. The new caboose is fine with default stats, but really the new express cars are too light. To make best use of them (ie: for overall game balance) they'll need tweaking of stats for express locos, largely for fuel consumption but also for other attributes.

Short version: you can use them with default stats but they'll make things a bit too easy.

Most remarkable perhaps is the Camelback's fuel costs, wow! Also further validation, I love the Eight Wheeler!

Yes, they're both good on fuel, which is an important consideration.

[RulerofRails]

The car weights in the uploaded express car pack start at 2 tons and are more or less about half the default ones. In your test year (1874), express weighs 2.6 tons. Default express for that era weighs 7 tons. Compare this to the default freight at 10 tons, and you can see why this isn't recommended for an accurate picture. In a haul-anything situation, 2 cars freight = 8 cars express.

This will help explain why the "express" engines (Eight Wheeler) and (Stirling) are performing so well. Their lack of Pulling Power isn't limiting them as much as normal. Obviously, balancing express engines for the new weights involves reducing their pulling power to bring them back down to sensible levels. At the moment most of your trains will hit top speed on flat ground while dragging 8 express cars.

If you want, check it for yourself, but I'm pretty sure about two slow-time hours equaling a normal day. So that's 60 slow time hours in a normal time month. I used slow time to get precise measurements. Since I made all the measurements there, I just kept that unit. I could have converted into normal days, but as it is I have enough data to get confused.

"Miles Per Hour" (mph) as used by the game, is according to my measurements: 2.35 miles* covered @ 1mph in one month. *(Miles as counted by the game's track mileage readout). So for 10mph we cover 23.5 miles in a month, etc.. Didn't see a way this makes sense, but thankfully things like this have never stopped us enjoying the game.

In terms of mileage per "game square", I always relied on the game's track mileage count. However, as an approximation, a small square crossed straight through is approximately 1.375 track miles. I went over my data (you will find this and water/oil level info on the RHS of this) and realized that I had mixed up track pieces and track mileage counts here as well. Anyway, all fixed now.

Re: your test data
For this bulk application the average speed count would definitely average out. My experiences were with a single engine, and let's just say that something goes funny with it over year end. I didn't try to figure out the details, but I lost trust in it. But, with a lot of engines that inaccuracy is probably averaged out close enough.

It sounds like you run even more locos than I ever did (my plays in the last year or so haven't typically seen excessive use of locos). I can see how you are wanting higher acceleration to clear jams. However, personally I don't buy a loco thinking "Oh, this one wont get into as many jams as the other." If I get a jam I tend to think that it's a problem with my track design/service placement. If the map isn't resource dense it's hard to get jams with good design.

Might be good in your test to record a count of the express loads hauled. I would comment that the Alder seems the odd one out. The Norris is only 1mph faster (8.3% faster), but hauls 233 more cars (49% more). If average speed is counted correctly, average speed being a by-product of acceleration, this can't be explained by lower acceleration level. Probably the speed was slow enough to result in a poorer balance of the supply/demand equilibrium? Unless you only ran the test over one year, and therefore trains that are already en route at the start of the test had a lot less chance of being able to complete a subsequent journey with a slightly reduced average speed?

[low_grade]

Hmmm, yes the Adler points to the need for a multi-year test perhaps, but otherwise loads/ave speed = 52 to 55 for all other locos, only 38 for the Adler, odd. Probably just a retest of the Adler hoping that randomness got in the way... This also points to a statistic to watch in future testing to give the data greater confidence. And I need to revert to the old express cars (I'm keeping the diner and caboose!) Of the various situations I'd like to test, this one falls into the mostly-flat-express-heavy-resource-dense category, for sure. Simple logic would say yeah acceleration and pax appeal for the win.

I have 39 stations, 89 trains, I don't know if that's a crazy ratio, but figuring most of my trains service 3 stations, that's approx 7 trains per station average, some stations I know are only serviced by 2 trains, and I'm sure others get maybe 15 total. So that's the jams.

Crashes also seem totally random, how did the Connie crash 4 times and the Stirling not once?!? So disable crashes. Breakdowns seem to match reliability much better. This is testing on 4 year old engines.

Also, you were estimating a post WWII loco might do 500 miles per year, but here most of the period B locos were doing over 500 miles per year, and the Eight Wheeler covered almost 700 miles, Adler still 340 miles.

Also (check my math here) but the percent chance for a breakdown per 1000 miles would be calculated by (breakdowns)*(1000/miles traveled)/(number of locos). The Ten Wheeler broke down 25 times, and traveled almost 600 miles, so per 1000 that would be 25*1000/600 = 42 breakdowns / 89 trains = 47% chance?!? Firefly worked out to over 60% chance, and all other trains were from 20% up to 37%. This seems rather high for 4 year old engines...

Also, using your cost calculator for a 4 year old Eight Wheeler travelling 680 miles per year with 52 ton consist (3 express at 2.6 and 3 freight at 10 plus I guessed 7 each for diner and caboose gives 801 maint vs 631 in the test, 3204 fuel vs 2378 in the test, and 13.9% breakdown vs 28.2% in the test...

Ah shoot just checking game effects I see pax is +50 and loco maint is -20... this brings loco maintenance up to 789, much closer to 801. The fuel is right if the consist only weighs 10 tons or if I only traveled 500 miles, or consist 30 miles 550.

Well, back to the drawing board, I can scrap the testing I've done so far, but I've learned a bit and will try try again...

[RulerofRails]

True, that mileage estimate is a bit conservative. It's only meant to be a starting point if someone is unsure what they are dealing with in their situation. I made some allowance for the average player not designing their service stops perfectly. Maybe it wasn't a good estimate, but I wanted to put something simple there to give a start. In the real world, mileage obviously varies. The Eight Wheeler is actually a fast train on flat ground. Dragging 80 tons (8 cars of freight) it will hit 79mph.

In terms of reliability, maybe your trains are running below the 60% Oil level? Also, the sheet only approximates the displayed breakdown chance in a specific circumstance. That's what I could calculate automatically. No calculation was made to make it a real "average". It becomes all technical and case specific depending on usage really fast, but maybe I should have made it clearer of not being an accurate average.

Gumboots reported inconsistencies in number of crashes as well. From what we have seen of the way that the game handles probability, there may be some discrepancy between the displayed chance, and actual chance. The other thing is that a few trains that are serviced poorly may make up a greater percentage of breakdowns and skew your statistics.

Diner you have is 2.6 tons (same as express), Caboose is 5.2 tons. So this brings your estimate for tonnage down by 6 tons. The other thing is are your
trains running full? The fuel formula is the one I have tested the most. I have little doubt that it's accurate for a given mileage and consist weight. I would recommend recording the amount of express hauled in your future tests. Here I posted the code that can be copied to your status page. For existing games since its LTD, you would need to make one recording of loads already hauled before the test and then just log them after each test run.

Yeah, quite a few scenarios use events that will affect loco profitability. The +50 on passengers is sure to favor the express engines.

[Gumboots]

Hmmm, yes the Adler points to the need for a multi-year test perhaps, but otherwise loads/ave speed = 52 to 55 for all other locos, only 38 for the Adler, odd.

The Adler is the weakest of the early locos. The Planet actually has more grunt, even though it's nominally more primitive.

And I need to revert to the old express cars (I'm keeping the diner and caboose!)

Use the defaults, not the WP&P cars. WP&P's are all over the place for weight.

Crashes also seem totally random, how did the Connie crash 4 times and the Stirling not once?!? So disable crashes. Breakdowns seem to match reliability much better.

Yup, that's what I found when I tested crashes and breakdowns.

Also (check my math here) but the percent chance for a breakdown per 1000 miles would be calculated by (breakdowns)*(1000/miles traveled)/(number of locos). The Ten Wheeler broke down 25 times, and traveled almost 600 miles, so per 1000 that would be 25*1000/600 = 42 breakdowns / 89 trains = 47% chance?!? Firefly worked out to over 60% chance, and all other trains were from 20% up to 37%. This seems rather high for 4 year old engines...

The way the reliability works is that breakdown chance increases rapidly for the first three years, then eases off to a shallow linear increase. So by the time your locos are 4 years old, they are almost as unreliable as when they are 8 or 10 years old. So yup, those figures sound about right if you are not using a caboose on all locos.

[low_grade]

Okay, let's try this again. Started with Southeast USA 1835, got rid of events, made myself filthy rich, built a double-tracked with stone bridges network connecting all cities, a few rough spots but mostly flat, no hotels, ran all 7+caboose auto consist. Ran for four years (reduced overhead by 90%) and added a few trains, upgraded some industries, got a save ready for 1839 and did a little A-era test with four year old locos in 1943, then took that save and advanced it to 1852 and did a B-era test of 4 year old locos in 1856.

88 engines total, not too congested around the bigger cities (I was more careful to keep junctions at least maybe 3 big squares away and service towers, too, I think this helped...) Worked out to be about 2/3 freight hauling, and ran at almost 100% capacity, but even a few locos hauling only 0-3 loads will drop the average, say 6.2 loads average, and given B period weights and Gumboots' 5 ton caboose that's 60 ton consists for the 2nd test. Used this and the data from maintenance and fuel costs to check RoR's formulas... oddly maintenance was much higher, but fuel while not always accurate was evenly +/- so on average the formula appears to be correct.

No big surprises but more confident data to represent the performance of the B-era locos at work on a map with lots of cargo and forgiving terrain. No crashes this time. Many of the breakdown figures are still way too high, and yes I'm using a caboose, and oil on average is at 70% I'd say.

Edit: realizing how many of these engines I had modded, I made another fresh 1.06 install (and added Gumboots' caboose) to also test the original 1.06 versions of the Connie, Duke, Ten Wheeler, American, Camelback, and Mogul. Also added a note with each of my modded locos to indicate how they differed from original. Interesting to note what may just be random variation, but may also be an indication of the cost of a breakdown, as now my Connie, which is faster (48mph vs 45mph) so really just a small difference from the original, with three more breakdowns ends up making a little less money. On the other hand, the original Duke, only a little slower than mine (50mph vs 55mph) had a bunch more breakdowns (randomly, I didn't fiddle with reliability) and ended up generating significantly less revenue.

[Gumboots]

Just bear in mind that, because it's an exercise in probability, you will only get a perfect match on breakdowns if you do an infinite number of trials. Doing just a few hundred trials will even the lumps out to some extent, but not completely.