Pier System Branch Pak128.Britian-Ex: Balencing vs Historic Accuracy

[PJMack]

In order to test the Pier_System branch in Simutrans-Extended, I had created a corresponding branch of Simutrans-pak128.britain for the purposes of testing as discussed elsewhere of the forum (https://forum.simutrans.com/index.php/topic,21164.0.html).  Such discussion if primarily on the coding and UI side of things, whereas here I would like to continue (or start anew) the discussion from the pakset designer's perspective of things to handle game balance, historic accuracy, accuracy to Britain, and where such conflict.

There are three discrepancies that come to mind with the existing elevated ways in the pakset.  Firstly, the axle load limits for the elevated ways are set well above that of any vehicles, making them ineffective.  Secondly, the costs associated with the elevated ways are significantly higher than that of the equivalent multi-span bridges, resulting in a "cheat" where such bridges are used for long straight sections.  The third issue is into date correctness of some of the viaduct types, specifically the iron girder elevated rail system, which has an intro date approximately the same as the first such system in New York City, and as far as I can tell, would not be introduced in Britain until the Liverpool Overhead.  Likewise, I cannot find any evidence of a steel overhead road system in Britain (though I could be wrong about that).

When creating the new pier system dat files, I used adjusted the costs of the bridges.  For the brick elevated pier set, the intro date is not enabled yet (due to a stone pier system not being completed yet), the axle load limits for such set to maximum, and the costs set to be slightly less than the equivalent bridges.  For the iron pier set, the intro date was set to 1860 (about the same existing), the axle load limits set to the lighter tank locomotives of the era (which happen to be about the axle load limit of the Manhattan elevated), the costs set to half the build and double the maintenance of the brick elevated pier system.

I am requesting any opinions, thoughts and recommendations on the proper adjustments for these as well as recommendations of new pier types.

[jamespetts]

I should note first of all that the current costs have not yet been fully balanced in the pakset. That is because full balance is still awaiting the balance critical features. In general, however, elevated ways and bridges should not cost different amounts: instead, the cost differential between bridges/elevated ways/piers and plain track/road should be the same cost differential as in reality. Even before we finalise the balance fully with the balance critical features, we can by way of interim balancing set up the relative cost differential between plain track/road and bridges/piers/[elevated ways] and tunnels to match reality, calibrating this to the level of the existing plain road/track cost. One thing that is notable at present is usage by players of elevated ways in the Bridgewater-Brunel server game far in excess of what would occur in reality, suggesting that these ways may be too cheap (or alternatively, transport generally may be too profitable, but that would require full balancing, including the balance critical features, to remedy).

As to introduction dates, generally, for infrastructure, the best introduction date would be the earliest date on which such infrastructure could realistically have been built, rather than the first date on which it was in fact built in the UK, so the current introduction date of the iron elevated way system would seem to be correct. As for steel overhead roads, a large system of these is unlikely for economic reasons, but it is likely that there will have been individual flyovers which in Simutrans cannot be represented as bridges (e.g. because of being diagonal).

Axle loads I believe have been discussed elsewhere, so I will not repeat my reply in that regard.

[PJMack]

Thank You.  This does clear things up a bit.

I have been doing a bit more research into British railway construction.  One resource I found helpful is Railway Construction, by William Hemmingway Mills (https://gutenberg.org/ebooks/50696) which does generally compare any costs as "these must always vary to a very large extent, according to the locality, facility of supply, and current prices of materials."  I did notice, looking at map, that brick viaducts are more common in the southern parts of Great Britain and stone more common in the north even at the same era.  (Geographic north vs south, I cannot pretend to understand where the cultural divide is.)  This implies that even within the island, there was likely variations in cost between such areas.  This would make balancing based on real costs rather interesting, as such cost ratios may not be the same say London vs Glasgow.

Another thing that may need consideration is the cost of building a viaduct vs an embankment.  I think a good balance point would be making it cheaper to build an embankment 2 tiles high for open countryside, but cheaper to build viaducts 2 tiles high in urban areas.  Once exception to this may be wooden trestles which were often placed temporarily before embankments could be built, and would have a very high maintenance cost which would encourage quick replacement if the line becomes permanent, and demolition if not.  Likewise for the iron girders, but with a little more initial cost than wood, and slightly lower maintenance.

I also observed many disused arch bridges, which leads me to believe that I the maintenance costs should be significantly lower than demolition costs.

[jamespetts]

Interesting, thank you for that. I have found some other interesting information there which I have added to this thread. We will have to standardise the costs to some extent - we may have to fix on an area in which there was much construction and normalise the prices for that area if we have sufficient detail in order to do so, although, looking through the book, there does not appear to be sufficient price detail in the book to allow for this, so we will have to search other sources in the thread to which I link above.

In terms of brick vs. stone, I would suggest calibrating costs on the basis that brick is the cheaper material. This is because there was a time before brick became widely available where masonry had to be used even in places where brick was later predominant. This would give players an incentive to build masonry structures in the era before brick and then switch to brick structures when they become available. Setting the costs the other way would give players an incentive always to build masonry structures and entirely to ignore brick structures.

I agree with your general conclusion regarding the relative cost of viaducts versus embankments, with the qualification that this applies to sloped sided embankments only. Straight sided embankments (i.e. those retained by retaining walls on both sides) should be more expensive than viaducts, albeit not necessarily very much more expensive.

As to the relative cost of maintenance and construction, I agree that we need to balance bridges and viaducts to take better account of this so that many of the structures that are cheaper to build, such as a wooden trestle, are more expensive to maintain. This is also likely to apply to infrastructure other than ways: wooden platforms, in particular, should be cheaper to build but more expensive to maintain than brick platforms. This will not necessarily apply universally: a long, pillarless structure (such as a lattice girder bridge) for crossing a long section of deep water may well be more expensive than a brick arch viaduct to construct, but also be more expensive to maintain.

We may also need to adjust the bridge building code so as to charge more for certain kinds of bridges (i.e. those with pillars) depending on the height of the pillars and whether they are built over water. This should then make bridges behave consistently with piers in this regard.

In terms of disused brick or masonry arch bridges, we cannot necessarily conclude that these cost more to maintain in working order than to demolish, as we do not know the structural soundness of these bridges and what load that they could carry in their present state. This may be something that it is worthwhile looking into further. Certainly, however, the maintenance cost of these structures is likely to be very low compared to metal or wooden bridges.

In terms of balancing more generally, there may be much to be said to working now towards the first attempt at balancing infrastructure costs relative to each other at least. This is because, at least once we have the pier system (and possibly the modification relating to the cost of bridges to which I refer above), we will have all the infrastructure specific balance critical features. The missing balance critical features will be relating to vehicles and inflation, but that does not stop us from balancing infrastructure cost, at least on an interim basis. This will have the advantage of improving the game balance (and disincentivising the excessive use of elevated ways that gives players an incentive to build networks in a way very different to that in which it would have been sensible to build them in reality), and also being the first testing ground for the planned use of real life costs for the calibration of balancing.

As we do not have any inflation mechanism in the game yet, we will need to normalise the costs to a particular year. I believe that the year 1900 is what I had planned to be the standard normalising point for all costs before adding an inflation mechanism.

Extracting all of the way related costs from this balancing data thread, we get as follows (using the Bank of England inflation calculator to normalise the prices):

Ship canal: £3,125/km in 1724; £5,581.58/km in 1900
The Tay Bridge: £197,824.30/km in 1871; £189,581.33/km in 1900
Railway maintenance (labour cost): £365.63 in 1946/km/year; £124.58/km/year in 1900 (1)
Early railway telegraph equipment: £467.14/signalbox in 1839; £394.28 in 1900 (2)
Wooden waggonways: £27.34/km in 1810; £17.47/km in 1900 (3)
Signalbox staffing costs: Small signalbox: £416/year in 1946; £141,75 in 1900; medium signalbox: £624/year in 1946; £212.62/year in 1900; large signalbox: £1,456/year in 1946; £496/12/year in 1900 (4)
Short bridge over railway:  £15,500/km in 1894; £16,390.80 in 1900 (5)
Railway depot: £28,000 in 1865; £28,622.22 in 1900
Maglev track costs: £9.1m/km in 2008 or £98,784.66/km in 1900 (6)
Railway depot maintenance: £6,000/year in 1931; £3,325.30 in 1900
Railway electrification 25kV overhead: £57,714.29/km in 1986; £1,375.92/km in 1900 (7)
Brick platform: £668.40/tile in 1932; £379,59 in 1900 ( 8)
Small station building: £33,000 in 1954; £7,351.09 in 1900
Small signalbox: £300 in 1909; £290.53 in 1900
Signal (with track circuits): £100/arm in 1909; £96.84/arm in 1900
Signal maintenance cost: £14-22/arm/year in 1905: £13.85 - £21.76 in 1900
Wooden trestle (rail) maintenance (single track): £56,373.39/km in 2013; £525.62/km in 1900 (9)
Railway construction (single track): £1,400.56/km in 1887; £1,498.74/km in 1900 (10)
Tram track maintenance: £468.43/km/year in 1935; £270.99/km/year in 1900 (11)
Tram track construction (55lb/yard): £2,345.52/km in 1881; £2,329.68 in 1900
Tram track construction (55lb/yard - excluding forge cost): £1,310.84/km in 1881; £1,296.90/km in 1900 (12)
Tramway electrification: £2,392/km in 1891; £2,472.63 in 1900
Tram track maintenance (98lb/yard): £234.30/km/year in 1892; £242.20/km/year in 1900
Tram track (47lb/yard): £1,864.51/km in 1877; £1,927.36/km in 1900
Tram track (upgrade of existing line): £1,750/km in 1878; £1,694.74 in 1900
Brick arch railway viaduct: £37,037.04/km (13)
Motorway: £9,320,206/km in 2011; £92,418.51/km in 1900 (14)
Motorway tunnel (assume concrete): £77,645,600/km in 2011; £769,928.35 in 1900 (15)
Motorway elevated section (assume concrete):  £42,650,400/km in 2011; £422,918.39 in 1900 (16)
Railway tunnel (Liverpool & Manchester Railway): £11,887.56/km in 1830; £10,396.93/km in 1900 (17)
Railway tunnel (Severn Tunnel): £71,347.03/km in 1873; £66,302.26 in 1900 (18)
Canal maintenance cost: £294.40/km in 1938; £161.21/km in 1900 (19)
Railway electrification (750v third rail): £7,770.52/km in 1925; £3,843.72/km in 1900 (20)
Railway electrification (25kV overhead lines): £12,432.83/km (nominal) in 1925; £6,149.66/km in 1900 (21)

(Costs are capital costs unless otherwise specified. All way costs would include the forge cost unless marked as renewal cost. Costs that are difficult to disentangle (e.g., station buildings from platforms) have been omitted)

Calculations

(1) Track maintenance: 5 platelayers per 3.2km of track each earning £4.50/week in 1946. This equates to £4.50 x 52 x 5 = £1,170/year for 3.2km, or £365.63/km in 1946
(2) Railway telegraph equipment: assume 7 signalboxes between and including London Paddington and West Drayton, cost between them of £3,270, so £467.14/signalbox.
(3) Wooden waggonways: £0.025/yard = 1093.6 yards/km x £0.025 = £27.34
(4) Signalbox staffing: Small box, assume £4/week for each signalman, 2x shifts of 8 hours per day, so 2x signalmen; £4 x 2 x 52 = £416/year. Medium box, assume £6/week, so £6 x 2 x 52 = £624/year. Large box, assume 2x signalmen on duty, one at £6 and one at £8 (so average of £7), 2x 8 hour shifts, so £7 x 2 x 2 x 52 = £1,456/year. Later power signalboxes not easy to calculate from these data.
(5) Assume double track railway, so 2x tiles, plus 2x tiles lead up, so total of 0.5km. Overall price was said to be £7,750, so £15,500/km
(6) AU$34m/km for dual track, equating to AU$17/km for single track, equating to £9.1m (conversion here).
(7) Romford to Upminster: 5.25km. £303,000 for whole length, so £57,714.29/km in 1986
( 8) Southend East station. Assume 5 tiles at £3,342, so £668.40/tile.
(9) Barmouth Bridge is 699m long and cost £39,405 to maintain in 2013. That equates to £56,373.39/km
(10) 1,428m at £2,000 is £1,400.56. This was a minor line, so assume 70lb/yard
(11) £936.86/route km. Assuming all routes double track, equating to £468.43/track km
(12) £2,370.70 per 1.808km = £1,310.84
(13) Hanwell viaduct - £40,000, 4x tracks (standard gauge), 270m long. £148,148.15/km for 4x tracks, or £37,037.04/km per line
(14) £17,045/yard. Divide by 2 as there are 2 lanes, to get £8,522.50/yrd. 1093.6 yards/km x £9,320,206/km
(15) £142,000/yard "" = £77,645,600/km
(16) £78,000/yard  "" = £42,650,400/km
(17) £47,788 cost; length = 2.01km https://en.wikipedia.org/wiki/Liverpool_and_Manchester_Railway. 2x track. Cost/track/km = £11,887.56
(18) £1,000,000 / 2 track / 7.008km = £71,347.03/km
(19) £184/mile = £294.40/km (at 1.6km/mile)
(20) £833,000 for 67 track miles = £7,770.52/km
(21) 40/25ths of the third rail cost (as only route miles are given):

[jamespetts]

I am in the process of undertaking some preliminary work on balancing adjustments on my pier_system branch of the pakset. I have started by disabling building elevated ways and bridges with supports over any buildings (and also adding the longstanding option to do this to simuconf.tab: it has been in the code for a long time, but does not seem to have made it into the actual simuconf.tab file).

I am going to see if I can do some infrastructure cost balancing on this branch, too, based on the above computations, which should assist in taking the first steps towards realistic balance overall. The vehicle based costs balancing will need to wait for more features, however.
Edit: I have added a spreadsheet with balancing data here.

[PJMack]

Thank you. 

I have come across some more details in the prior mentioned book to note.  One being that embankments were used up to 25-30 feet connecting to viaducts over land.  If I am not mistaken, that is one or two Simutrans tile heights.  For tunnels, It notes that building a 50 to 70 ft trench may been cheaper than building a tunnel, and that tunnel costs are £32 to £150 per linear yard (which equates to about £30,000 to £145,000 per km in 1900) depending on conditions which are usually somewhat unknown at the time of construction.  (Brings to mind the extreme example of Boston's Big Dig.)  I assume that this is for two line track, so we may need to half these values.

One thing that may need consideration when balancing is the differences in reality and Simutrans that are forced upon the player.  One example is the dual scale that station incur where to have realistic platform an train length involved stations with 2+km platforms and 125m distance between each one.  This is not currently correctable without greatly limiting map scale, so it is probably best to work around the situation.  The other more relevant example is that it take two tiles for a dual track line, but only one for a city road.  This means that for elevated ways (such as the Liverpool overhead) that only one direction could be placed above the road and placing the other would necessitate the removal of buildings not represented in the real world.  Changing this would be an immense coding project, it is probably also best to worked with the situation.  I would also be interested in other peoples take on this.

(Just a side note, in addition to inflation and labor costs, when the time comes, it may be worth considering adding some building material costs as well to the financial equations)

[jamespetts]

Excellent, thank you for that. I should note that, following on from the above discussion of balance calibration, I have started a more general discussion about balance calibration here, and I should be interested in your input. This thread can be retained for discussion of issues specific to elevated way supports/piers.

Embankments

Thank you for that - that is helpful. I would suggest that the landscaping cost be calibrated, as you suggest, to be lower than the viaduct cost for a sloped sided embankment of 2 tiles high or less.

Calibration

This is largely addressed in the long balancing thread above. Platform lengths are not, and this is something that will need to be considered more generally. Platform costs may have to be reduced to the graphical scale of the vehicles (30m/tile, I believe, if I recall correctly) by linear length rather than the general distance scale.
Dealing with the width issue for towns is complex, and there is not an obvious solution to this that would not cause anomalies elsewhere that exceed in effect the anomaly caused by not addressing this at all. I should be interested in people's views on this question.

Inflation

This is probably better discussed in the general balance thread linked above - what categories that we have may need to depend on what historical material that we can find. If anyone knows of any good sources for historical information on labour and commodity prices going back at least 200 (and preferably 350) years, that would be very helpful.

[PJMack]

I have just pushed an update for the 'L" style wrought iron piers.  They can now be built automatically.  I have not yet set the costs and the into date is still a little early.  One general question I have for the community is whether or not to make such piers rail only and have a similar one for roads, as the intro date for that style is different for rail (1860s New York) and road (1920s Chicago?).  Both were eventually used in Britain (Liverpool overhead, a few road "flyovers").  I also did not put any railings on that style pier as for rail, they would not have had any, and for road, the railings would be far too small to show up in (about 42" tall maximum, about 2 pixels).

[jamespetts]

Excellent, thank you for this. I will test this when the merge issue with the master branch has been resolved, as discussed on the other thread.

As to the iron elevated way, whether to have a separate one for rail and road really depends on the reason for the difference: was there something inherent about railways that meant that a type designed for road usage was fundamentally incompatible? If so, having a separation would make sense. Was the issue more that the early types could not handle the weight of rail vehicles, or were too expensive to be worthwhile for roadways? In which case, this would not justify a separation of types, and we should instead have these weight/cost limitations embedded into the single structure.

Incidentally, the most significant elevated way support type so far missing from the new system is the wooden trestle - is that something that you had planned to add?

Thank you again for your work on this.

[PJMack]

As to the iron elevated way, whether to have a separate one for rail and road really depends on the reason for the difference: was there something inherent about railways that meant that a type designed for road usage was fundamentally incompatible? If so, having a separation would make sense. If so, having a separation would make sense. Was the issue more that the early types could not handle the weight of rail vehicles, or were too expensive to be worthwhile for roadways? In which case, this would not justify a separation of types, and we should instead have these weight/cost limitations embedded into the single structure.

For some of the rail type, there would not actually be a deck, but the ties (sleepers) laid directly on a structure no wider than track gauge.  Other types would have a deck, but depending upon how they were built, may still have been too narrow for roads.  The type I included did have a deck wide enough for the in game rail and road types (though the dirt and cobble roads may need to be narrowed by a couple pixels).  As for the delay between rail and road usage, it could have simply been lack of demand for the time period.  Elevated road systems over other roads would not have been needed before cars became popular. 

Incidentally, the most significant elevated way support type so far missing from the new system is the wooden trestle - is that something that you had planned to add?

I did start a wooden trestle blend file but was having trouble getting the proportions right so I put it on the back burner.  I am currently working on concrete elevated ways, but in the meantime, I did have a question about the wooden trestles.  Where there ever switches (points) on wooden trestles?  I have not been able to find any, but as there are not very many wooden trestles left, and not much documentation from the time, I could be missing something. 

[jamespetts]

I suspect that the only thing stopping iron being used for elevated roadways would have been lack of demand, as you indicated: indeed, I believe that there was an iron elevated roadway inside Paddington station for a long time. The remains of it are still there, although it is now disused (although I remember it being used for taxis when I was younger).

As to the other types, I had forgotten about concrete: this is even more important than the wooden type. I think that, once we have the wooden and concrete types, and some aqueducts, we will have a complete enough range to be able to replace the existing elevated ways and some bridges with this system.

[PJMack]

I completed the graphics for the concrete elevated ways and have pushed them.  The intro date has not yet been set and the prices not yet finalized.  For the prices, the modern concrete elevated structures appear to be four times as much as their history counterparts bases on the discussion below and in the other forum thread.  I am not sure if this is the right thing for game play.

For aqueducts, I have found aqueducts with slight curves, however have not found any evidence of waterways branching or crossing on aqueducts, nor any evidence of locks on aqueducts, so am not planning on supporting those.  For barge canals, I am planning on requiring them to be three levels above a track or public road and two levels above an owned road, as to deal with the depth of the water.  For narrow boats, it is questionable whether that depth needs to be taken into account.  (For the drawing scale, each tile is 30m x 30m x 6m).

[PJMack]

I have completed basic aqueducts for stone and concrete.  They both allow for tub, narrow, and barge canals, thus requiring an extra tile between the top of the water line and the top of any way below.  They also do not permit intersections.  For the stone aqueduct to work properly, I had to fix a bug in the auto building feature of the piers.

I did some more research and did find that the River Card Aqueduct was converted into a rail viaduct.  Originally, the depth of the canal line (Glasgow, Paisley and Johnstone Canal) had a quaysides with a 6ft depth for the flyboats being operated.  The New Semington Aqueduct also has a 2 meter depth.  (These are according to wikipedia articles.)  I will put cost details in the appropriate forum thread, but they appear to be similar to that of a rail viaduct.  This implies that for the stone pier system, I could allow narrow-boats and tub boats on the ordinary ones, however I would need adjust the width of the tow paths by a couple of pixels.

[PJMack]

I have completed the timber piers, however am having trouble figuring out what the costs should be.  I did find ample studies on modern timber bridges (https://intrans.iastate.edu/app/uploads/2018/08/fplrp593.pdf, https://www.fpl.fs.fed.us/documnts/pdf1996/sowar96a.pdf, https://www.fpl.fs.fed.us/documnts/fplgtr/fplgtr125.pdf, https://babel.hathitrust.org/cgi/pt?id=umn.31951d00276578k&view=1up&seq=1), however those are all for pressure treated structures in the modern day US where the economics and bridge construction is rather different.  (The most common type of bridge near me has a steel span and concrete columns, wooden rail ties are still used, I have never seen a brick bridge in person, etc.) 

I looked into the Usk Bridge again.  It did burn down when nearly complete, so the cost may or may not include building the bridge twice.

On thing I did read is that early wooden bridges needed to be rebuild every twenty to thirty years, which would average the maintenance to at least 1/30th the construction cost (it would fluctuate).  Given the difference in simutrans capital vs maintenance cost conversions, I temporarily set the wood pier construction cost low, with maintenance cost half the construction cost.

[PJMack]

I have complete the tentative costs of piers and bridges for this branch as summarized below:

All construction and maintenance costs are based on two tile high, double track. 

For brick and stone bridges, I used the rounded median construction costs of ~£90k and £100k per km respectively.  As no maintenance details, I had to approximate them using an estimated lifetime.  Synchronizing the brick and stone maintenance costs is not too unrealistic and prevents the unrealist incentives from players replacing old stone piers with brick ones.

For the "El" type piers, lacking reliable sources for Britain, I used the information available from Handbook of cost data for contractors and engineers by Halbert Powers Gillette, second edition, 1910.  It contained information for both construction and maintenance costs.

For the wood pier types, I used the same resource as the "El", only I did not have any maintenance costs, but the average lifetime of the components.  Unfortunately due to the capital to recurring cost ratios, the wood pier monthly maintenance costs are equal to the construction costs.

For Concrete piers and the iron viaducts, the costs and maintenance were set for game balance as the few real-world costs I could find were anomalous. 

For bridges matching a pier type, costs and maintenance were set to be the same as their corresponding pier type, and weight and speed limits adjusted accordingly.  All other bridges had their costs and maintenance multiplied by 6 to be in the range of the piers' costs and maintenance.

[PJMack]

After a bit of game play, I realized that the maintenance costs were not balanced properly.  I realized I had used the short-term cost (GBP to simucents) conversion factor for maintenance rather than the long-term conversion factor.  I pushed the correction to this.  Now the wood pier monthly costs are more reasonable and time-span where the total costs (construction plus cumulative maintenance costs) of a stone pier verses a wooden one is now 46 years instead of 7 months.

[jamespetts]

As set out in the thread on the major projects forum, I have now incorporated this. As a result, I have incremented the version number in the pakset to 0.9.4, the first version increase in quite some time. Thank you for your excellent work on this.

There may yet need to be some further work in calibrating/balancing way related costs, but this is a very good starting point, and makes the cost of elevated ways much easier to control and much more sensible.