RAGL through the seasons

Discussion of the change in game-play over the various seasons of RAGL.

Announcements and discussion about community-run events.
TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

Thanks for the request to look into game length Jacko. While getting started on that I got side tracked looking at the chat logs. There are lots of conventions in RAGL, but one of the most contentious is whether to say "glhf" at the start. The idea of wishing someone luck when they are playing against you seems a dubious strategy, but perhaps this is cancelled somewhat by persuading them to "have fun", rather than play properly. The raw numbers look like this:

Code: Select all

Season  glhf   hf    gl   hfgl
   1   36.4%  2.6%  1.2%  0.2%
   2   35.2%  1.7%  4.0%  0.0%
   3   38.0%  1.1%  3.6%  0.0%
   4   45.2%  1.7%  2.7%  0.5%
   5   42.9%  1.2%  6.0%  0.2%
   6   46.6%  0.7%  9.4%  0.0%
   7   44.4%  1.2%  9.1%  0.0%
   8   50.7%  1.0%  5.5%  0.0%
   9   58.6%  3.7%  4.5%  2.5%
Looking at the rates which these wishes were made we can see that "glhf" is indeed the standard, and has steadily grown in popularity over the years. The number of players who want their matches purely dependent on skill has held fairly steady - somewhere between one and two percent. In Seasons 1 we saw a slightly higher rate, and in Season 9 a record-breaking 3.7%. Players wishing just "gl" are presumably very confident in their own ability, and this saw a peak in Seasons 5 and 6. For completeness I included the stats for "hfgl" - the choice of those players trying to confuse their opponents in the early game.

So what about the win rates? I've looked at games where only one player gives a particular message, and calculated the win rate in just those games.
013_RallyWinRate.png
013_RallyWinRate.png (38.1 KiB) Viewed 15247 times
Due to the low number of games represented by some of these data points, there's a lot of error in the results. However the "glhf" line is backed up by a reasonable amount of games (e.g. 92 wins from 150 games in Season 9). This shows that players saying "glhf" have done significantly better than those who don't. Perhaps this is because very new players don't know about this convention, or perhaps because players who are panicking fail to type it correctly!

I also investigated whether saying glhf more times makes a difference. This was fairly inconclusive, but it did throw up the following game between Bain and Admiral Mo which contains the most "glhf"s by the players:

Code: Select all

RAGL-S08-MASTER-GROUP-AMO-BBB-G2.orarep abridged chat

Bain: glhf
Admiral mo: glhf!
Bain: glhf
[Bain sets correct options]
Admiral mo: glhf
[Bain kicks a load of specs]
Battlefield Control: The game has started.
Bain: glhf!
^^ZxGanon^^|RV (Spectator): hf
Radical Centrist (Spectator): gl hf
Antarctica (Spectator): raglhf
...
Bain: GG
Admiral mo: gg
Antarctica (Spectator): gg
bµg: gg
Bain: gg wp
^^ZxGanon^^|RV (Spectator): gg
Admiral mo: you too!
goatkeeper (Spectator): gg
This brings us on to another very common message. This time I looked at the win rate of the player who says "gg" first:
013_GGWinRate.png
013_GGWinRate.png (22.61 KiB) Viewed 15295 times
We can clearly see that saying "gg" first has never been a good tactic, but over the years it has got steadily worse.
Last edited by TTTPPP on Tue Aug 11, 2020 6:24 pm, edited 2 times in total.

lawANDorder
Posts: 140
Joined: Tue Oct 24, 2017 3:20 pm

Re: RAGL through the seasons

Post by lawANDorder »

> or perhaps because players who are panicking fail to type it correctly!

Rumors say some players even program a glhf-hotkey! :lol:

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

So an excursion into game lengths, and then I'll get back to looking at the infantry queue. First - the average and maximum game lengths against season:
014_GameLength.png
014_GameLength.png (32.38 KiB) Viewed 15244 times
The graph clearly shows that the longest games have tended to get shorter over the seasons, but also the average game length has dropped too. In Season 1 it was almost 16 minutes, and in Season 9 it was at a new low of 12 and a half.

When compiling these stats I excluded games that hadn't ended with a clear winner and loser. Some games included pauses, but this didn't have a big impact. The longest games were as follows (and in fact none of them contained pauses):

Code: Select all

Season 1: RAGL-S01-MINION-R09-TRX-ABC.orarep 68.5 minutes
Season 2: RAGL-S02-MASTER-R02-ABC-BRF.orarep 87.3 minutes
Season 3: RAGL-S03-MINION-R05-TTT-LUC.orarep 51.6 minutes
Season 4: RAGL-S04-RECRUIT-R06-SCR-DRG.orarep 68.3 minutes
Season 5: RAGL-S05-RECRUITB-R05-CRD-LAK.orarep 45.3 minutes
Season 6: RAGL-S06-MASTER-R03-UPS-UNO-G2.orarep 51.2 minutes
Season 7: RAGL-S07-MASTER-GROUP-ANJ-NNG-G1.orarep 35.7 minutes
Season 8: RAGL-S08-MASTER-GROUP-BUG-AMO-G2.orarep 47.7 minutes
Season 9: RAGL-S09-MASTER-GROUP-MRC-AMO-G2.orarep 45.4 minutes
Next I wanted to look at how the distribution of game lengths have changed over time.
014_GameLengthDistribution.png
014_GameLengthDistribution.png (46.29 KiB) Viewed 15244 times
The graph doesn't really show an obvious change in game lengths, but it does show that most games finish between 5 and 20 minutes.

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

I had a few more thoughts about factors affecting game lengths. First I noticed that the list of the longest matches had a bias towards Masters games. I calculated the average game length for matches that specified which division they were from in the filename:
015_GameLengthDivision.png
015_GameLengthDivision.png (35.32 KiB) Viewed 15232 times
Overall it's possible that Masters games last slightly longer on average, but it's certainly not a consistent trend.

The other thing I wondered about was the notion of stalematey Allies mirror matches. I found the average game lengths for Allies and Soviet mirror matches, and I split the other games into those where Allies won and those where Soviets won:
015_GameLengthFactions.png
015_GameLengthFactions.png (43.27 KiB) Viewed 15232 times
I was quite surprised by this - it seems like in recent times then Soviet matches have lasted slightly longer on average. This was backed up by the list of the longest matches - for each I looked up the factions involved and found that there were actually no Allies mirror matches at all.

Code: Select all

Season 1: RAGL-S01-MINION-R09-TRX-ABC.orarep 68.5 minutes (Soviet v Soviet)
Season 2: RAGL-S02-MASTER-R02-ABC-BRF.orarep 87.3 minutes (Soviet v Allies)
Season 3: RAGL-S03-MINION-R05-TTT-LUC.orarep 51.6 minutes (Soviet v Allies)
Season 4: RAGL-S04-RECRUIT-R06-SCR-DRG.orarep 68.3 minutes (Allies v Soviet)
Season 5: RAGL-S05-RECRUITB-R05-CRD-LAK.orarep 45.3 minutes (Soviet v Soviet)
Season 6: RAGL-S06-MASTER-R03-UPS-UNO-G2.orarep 51.2 minutes (Allies v Soviet)
Season 7: RAGL-S07-MASTER-GROUP-ANJ-NNG-G1.orarep 35.7 minutes (Soviet v Soviet)
Season 8: RAGL-S08-MASTER-GROUP-BUG-AMO-G2.orarep 47.7 minutes (Soviet v Soviet)
Season 9: RAGL-S09-MASTER-GROUP-MRC-AMO-G2.orarep 45.4 minutes (Allies v Soviet)

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

So back to the infantry queue. I wanted to look at spies, and what sort of value the English spy provides.
016_SpyUsage.png
016_SpyUsage.png (52.17 KiB) Viewed 15189 times
Spies appear in about 10% of games, and slightly more often from English players than the other Allies. In Season 6 there was an oddity because no English player queued a single spy. On the whole then spy usage has fallen since Season 1, but has remained fairly steady since about Season 3 (except for an anomolous dip in Season 6).

If instead we consider the average number of spies build by players building at least one spy, then we get a graph like this:
016_SpyCount.png
016_SpyCount.png (53.16 KiB) Viewed 15189 times
Again we can see that English players use more spies on average than the other Allies. Also, comparing with the last graph, France uses spies more often than Germany, but queues fewer of them per game.

I should also give an honourable mention to SirCakealot, as I have completely ignored his two spies built when playing as Russia (RAGL-S03-MINION-R04-BLK-SIR.orarep). This was the only occasion a Soviet player obtained double tech and used it to build spies.

The value of English spies is limited by the quantity of spies that can usefully be built in a game. I compiled a list of all games where a player queued ten or more spies.

Code: Select all

33 spies by Medium Tank  (Germany) in RAGL-S02-MASTER-R07-MET-ABC.orarep
30 spies by Blackened    (England) in RAGL-S08-MINIONA-GROUP-BLK-PIE-G2.orarep
25 spies by Gatekeeper   (England) in RAGL-S03-MASTER-R02-GAT-FAR.orarep
15 spies by Unano        (Germany) in RAGL-S06-MASTER-R03-UPS-UNO-G2.orarep
14 spies by Bamsi        (England) in RAGL-S04-RECRUITA-R06-LOF-BAM.orarep
14 spies by SeaCucumber  (England) in RAGL-S04-RECRUIT-R06-SCR-DRG.orarep
13 spies by Mr Bain      (England) in RAGL-S08-MASTER-GROUP-UPS-BBB-G1.orarep
12 spies by NoobMapMaker (Germany) in RAGL-S01-MINION-R07-JAY-NMM.orarep
12 spies by Firekeeper   (England) in RAGL-S02-MASTER-R07-FIV-GAT.orarep
11 spies by kyrylo       (England) in RAGL-S01-MASTER-R01-KYR-MTR.orarep
11 spies by Barf         (England) in RAGL-S02-MASTER-TIEBREAK-BRF-GAT-1.orarep
11 spies by Barf         (England) in RAGL-S02-MASTER-R06-FIV-BRF.orarep
11 spies by Craig Christ (England) in RAGL-S04-RECRUITB-R06-LCN-CCT.orarep
11 spies by SeaCucumber  (England) in RAGL-S04-RECRUIT-R02-SCR-MAN.orarep
11 spies by SeaCucumber  (England) in RAGL-S04-RECRUIT-R01-SCR-MAN.orarep
10 spies by 3. Lucian    (England) in RAGL-S04-RECRUIT-R04-LCN-MCV.ORAREP
10 spies by mechANIC     (England) in RAGL-S04-RECRUIT-R04-LCN-MCV.ORAREP
10 spies by Bamsi        (England) in RAGL-S04-RECRUITA-R02-S84-BAM.orarep
10 spies by Bain         (England) in RAGL-S09-MASTER-GROUP-MRC-AMO-G2.orarep
We can see a very high proportion of players using England here, although the most spies queued was by Medium Tank as Germany. In this game Medium Tank apparently queued 33 spies, which seemed a bit insane, so I investigated further. Sadly the replay was too old for me to watch, but the list of actions involving spies looks something like this:

Code: Select all

Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
Start production of 5 spies
[Some other stuff]
Start production of 5 spies
Start production of 5 spies
[Lots of other stuff]
Pause production of a spy
Pause production of a spy
Pause production of a spy
Cancel production of 14 spies
Cancel production of 14 spies
Cancel production of 14 spies
I'm not 100% sure I've interpretted the file correctly, but it does look very much like the spies were queued by accident, and then they were cancelled once MT realised.

The second entry in the list is by Blackened in Season 8, so I was able to watch the replay. In this case Blackened queued thirty spies right near the end of the game but only two hit the field and neither made it near the enemy base.

The third entry (Gatekeeper in Season 3) seems very similar - 25 spies queued right near the end of the game.

These go to show the limitations of looking at the number of units queued. When a player queues a lot of an unusual unit then it is often a mistake or inconsequential to the outcome of the game. It would be better to instead somehow add instrumentation to the official replay client that is able to output the events that actually happened in the game.

The fourth entry in the list is from Unano, and he actually does build a lot of spies. However it's probably worth keeping this in the contex of the rest of his build - single barracks into Tech Center, Tanya (immediately sniped) and a bunch of spies (most get a disguise, but Upps spots them anyway), twin gap generators and a wall of camo pillboxes. He also throws in a lategame transition into light tank spam and crushes two of his disguised spies by chronoshifting enemy tanks on top of them. Fifteen spies are built and they are all killed off in some way or another without generating any value.
016_SpyUnanoVsUpps.png
016_SpyUnanoVsUpps.png (123.55 KiB) Viewed 15189 times
So even in this extreme case, if Unano had been England then he would have only saved himself $3750 on spies - that's about 20 rifles and 6 rockets (or 4 light tanks and a gap generator) - and he wouldn't have been able to pull off the same stunts using the German Chronosphere.

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

I thought back to the rifle-rocket ratio post from earlier, and wondered if there was a noticable change in the ratio as the game went on. Here I've plotted a line for each season. It's a bit difficult to interpret this graph, so the S01 point at (1-50, 3) means that the average r-r ratio for the first 50 infantry during games in Season 1 was 3. The next point on the same line is roughly at (51-100, 2) which means that the average r-r ratio for the next 50 infantry during those games was 2.
017_RifleRocketRatioThroughGame.png
017_RifleRocketRatioThroughGame.png (50.58 KiB) Viewed 15112 times
Other than explaining how to read it, there's not a huge amount to talk about for this graph. In every season then the rifle count starts high, drops and then levels off. This makes sense because it's very unusual to see rockets on the field in the first batch of units. Seasons 1 and 2 are noticable outliers with far more rockets per rifle, but otherwise they follow the same trend.

This got me wondering about other units, and how the proportions change during the game. I've plotted these on a logarithmic scale to try to make the graph a bit more readable. Each line in this graph is the average over all nine seasons of RAGL.
017_UnitProportionThroughGame.png
017_UnitProportionThroughGame.png (70.04 KiB) Viewed 15112 times
At the top we have rifles and rockets - between them making up >95% of the infantry. As discussed then the rifle percentage drops during the game, as the rocket density increases.

Engineers, grens and dogs are all early game units. Players perhaps make one or two in the first 50 units, but after that they're very rare. Perhaps it's experimental error, but all three seem to have a very slight uptick in popularity after the first 300 units.

Most of the other units stay fairly level. Tanya and shock troopers are obviously more likely to be seen in the late game since they require a Tech Center. Shock troopers can only be build by Russia, so when they reach 2-3% of all units that translates to something like 10-15% of all Russian units.

Flamers have a very slight increase in popularity as the game goes on, which is possibly because they're great at cleaning up buildings to close out a game. Medics conversely have a very slight decrease in popularity during the game. Perhaps this is because there are more units that one-shot infantry later on (artis, V2s, Tanya, shockies, etc.).

User avatar
Punsho
Posts: 143
Joined: Wed Jul 18, 2018 2:56 pm
Location: Lithuania

Re: RAGL through the seasons

Post by Punsho »

It would be quite interesting to see in how many games tech play mattered. One way to do that I think would be to see how long did it take to win after building a dome / tech centre, and probably a separate graph for losses. Graph could probably show how many games were played without tech, how many ended in 0-2 minutes after building dome, 2-4, 4-∞, how many ended in 0-2 minutes after building tech centre, 2-4, 4-∞. I don't know the best numbers, but I think if a game ended after 2-3 minutes after building a tech structure then I don't think that the tech structure changed the outcome of the game and was only used to help finish it

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

Finishing off this series of posts about infantry let's look at the players who've used (queued) the most of each unit in a game. Here's the list, and I've included the game duration too:

Code: Select all

Rifle    1920 morkel (RAGL-S09-MASTER-GROUP-MRK-ILM-G1.orarep - 38 minutes)
Rocket    782 morkel (RAGL-S09-MASTER-GROUP-MRK-ILM-G1.orarep - 38 minutes)
Engineer   16 WaRRyKeR (RAGL-S04-RECRUITC-R08-XY-WKR.orarep - 24 minutes)
Gren       59 SirCakealot (RAGL-S06-MINION-R01-JJK-SIR-G2.orarep - 19 minutes)
Flamer    115 netnazgul (RAGL-S05-MINION-R02-NNG-SOT-G2.orarep - 37 minutes)
Dog        20 dragunoff (RAGL-S04-RECRUIT-R06-SCR-DRG.orarep - 68 minutes)
Thief     125 Booby (RAGL-S03-RECRUIT-R11-BOO-OTQ.orarep - 22 minutes)
Shockie   281 Anykeyich (RAGL-S04-RECRUITC-R06-ANY-WKR.orarep.orarep - 20 minutes)
Medic     138 lucassss (RAGL-S03-MINION-R10-LUC-KUB.orarep - 17 minutes)
Mechanic   36 Kernel Panic (RAGL-S09-MINION-GROUP-KNL-ICE-G2.orarep - 28 minutes)
Spy        33 Medium Tank (RAGL-S02-MASTER-R07-MET-ABC.orarep - 30 minutes)
Tanya      39 SeaCucumber (RAGL-S04-RECRUIT-R06-SCR-DRG.orarep - 68 minutes)
We already looked at spies, and saw that just because someone queues a lot of a unit, it doesn't mean they're going to produce them. Other factors also need to be considered, such as the game ending before the units are produced.

Rifles and rockets form the core of an army and we saw before that the number queued has been creeping up over the seasons. Morkel holds the record for the most of both queued in a game, and it was an incredible game. Five Aces casted this game here: https://www.youtube.com/watch?v=zK-EglXyuBM
018_infantryMorkelVsILikeMenFiveAces.png
018_infantryMorkelVsILikeMenFiveAces.png (193.01 KiB) Viewed 14959 times
This queuing is a little ridiculous though because even at maximum production speed (rifle=1.2s, rocket=3.6s) then this queue would take 85 minutes to get through, and the game finished after 38. For comparison then ILM only queued 976 rifles and 372 rockets - the equivalent of about 42 minutes.

The most Tanyas queued in a game is in a really long game, which makes sense, but 39 still sounded like an awful lot. It turns out that this is because SeaCucumber was using shift-queuing which is niavely interpretted by my script as adding 5 Tanyas at once. The game engine only allows a single Tanya to be built at a time though, so there are more like 10-20 Tanyas in the game.

The highest flamer count is from a game between Netnazgul and Sotsch. Net goes for a 10 rifle, 5 rocket, 5 flamer mix of infantry for most of the game with mixed success. You can watch the replay via SoScared stream - it's a really nice game with most of the Soviet tech tree on display. Game 2 starts at about 17 minutes: https://www.twitch.tv/videos/251359565
018_infantryNetnazgulVsSotsch.png
018_infantryNetnazgulVsSotsch.png (46.65 KiB) Viewed 14959 times
The player with the most (queued) mechanics was Kernel Panic in Season 9 (RAGL-S09-MINION-GROUP-KNL-ICE-G2.orarep). Kernel Panic manages to grab something like nine heavies and seven mammoths, as well as reviving several of his ore trucks. Unfortunately this was one of the few S09 Minions matches that Biomecaman didn't cast, however you can watch a similar strategy by Kernel Panic against Duke Bones here: https://www.youtube.com/watch?v=CqZ1JqHIm2A (Beware the global rachet sound effect is quite obnoxious after a while).
018_infantryKernelPanicVsIcesinithol.png
018_infantryKernelPanicVsIcesinithol.png (268.35 KiB) Viewed 14959 times

SirCake
Posts: 393
Joined: Thu Feb 04, 2016 5:40 pm

Re: RAGL through the seasons

Post by SirCake »

Phah!.. Fifty-Nine grens in a game, what a paltry ammount!

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

The 59 grens game was a great effort SirCake! I didn't write more about it because I couldn't really tell if it was a good idea and I don't think there's a cast to link to, but it was fun to watch.

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

The next queue is the vehicles. We have all the same caveats for vehicles as we do for infantry - the replay format doesn't tell us whether the unit is produced, only whether it is added to the queue. Disclaimers out of the way, the raw numbers for each season look like this:

Code: Select all

SEASON 1
Arti: 3104 (215 builds (42%), 49% wins)
Medium Tank: 2801 (198 builds (39%), 58% wins)
Heavy Tank: 1772 (136 builds (27%), 54% wins)
Flak: 1290 (123 builds (24%), 48% wins)
Harvester: 1281 (427 builds (84%), 49% wins)
APC: 896 (156 builds (31%), 48% wins)
Light Tank: 878 (164 builds (32%), 48% wins)
MCV: 686 (395 builds (78%), 52% wins)
V2: 640 (94 builds (19%), 48% wins)
Ranger: 314 (197 builds (39%), 53% wins)
Tesla Tank: 238 (25 builds (5%), 65% wins)
Mammoth Tank: 51 (6 builds (1%), 40% wins)
Demo: 30 (11 builds (2%), 55% wins)
Radar Jammer: 18 (10 builds (2%), 80% wins)
Phase Transport: 16 (4 builds (1%), 33% wins)
Minelayer: 15 (13 builds (3%), 50% wins)
Mobile Gap Generator: 13 (8 builds (2%), 50% wins)
Supply Truck: 5 (5 builds (1%), 20% wins)

SEASON 2
Medium Tank: 4392 (286 builds (48%), 62% wins)
Arti: 2421 (243 builds (41%), 53% wins)
Harvester: 1605 (533 builds (89%), 50% wins)
Heavy Tank: 1436 (122 builds (20%), 48% wins)
Light Tank: 1429 (217 builds (36%), 49% wins)
Flak: 918 (116 builds (19%), 46% wins)
MCV: 854 (483 builds (81%), 54% wins)
Ranger: 551 (310 builds (52%), 54% wins)
APC: 551 (142 builds (24%), 43% wins)
V2: 511 (91 builds (15%), 48% wins)
Chrono Tank: 87 (8 builds (1%), 88% wins)
Mammoth Tank: 65 (8 builds (1%), 75% wins)
Tesla Tank: 55 (5 builds (1%), 80% wins)
Phase Transport: 31 (14 builds (2%), 50% wins)
Radar Jammer: 30 (13 builds (2%), 62% wins)
Minelayer: 27 (15 builds (3%), 80% wins)
Demo: 24 (9 builds (2%), 67% wins)
Mad Tank: 12 (3 builds (1%), 100% wins)
Mobile Gap Generator: 10 (5 builds (1%), 100% wins)
Supply Truck: 4 (4 builds (1%), 75% wins)

SEASON 3
Medium Tank: 4540 (259 builds (40%), 58% wins)
Arti: 1993 (172 builds (27%), 48% wins)
Light Tank: 1969 (200 builds (31%), 47% wins)
Heavy Tank: 1682 (159 builds (25%), 55% wins)
Harvester: 1555 (570 builds (89%), 52% wins)
Flak: 1131 (165 builds (26%), 54% wins)
APC: 1099 (186 builds (29%), 47% wins)
MCV: 1042 (523 builds (82%), 55% wins)
V2: 990 (107 builds (17%), 50% wins)
Ranger: 378 (266 builds (42%), 48% wins)
Mammoth Tank: 170 (17 builds (3%), 88% wins)
Mad Tank: 95 (2 builds (0%), 100% wins)
Chrono Tank: 54 (4 builds (1%), 75% wins)
Minelayer: 28 (16 builds (2%), 50% wins)
Phase Transport: 18 (10 builds (2%), 20% wins)
Tesla Tank: 11 (4 builds (1%), 50% wins)
Demo: 10 (8 builds (1%), 50% wins)
Radar Jammer: 4 (4 builds (1%), 75% wins)
Mobile Gap Generator: 4 (3 builds (0%), 100% wins)
Supply Truck: 1 (1 builds (0%), 0% wins)

SEASON 4
Medium Tank: 5766 (341 builds (35%), 52% wins)
Heavy Tank: 4380 (397 builds (40%), 56% wins)
Arti: 3629 (259 builds (26%), 46% wins)
V2: 2743 (341 builds (35%), 55% wins)
Flak: 2698 (360 builds (36%), 54% wins)
Harvester: 2393 (896 builds (91%), 52% wins)
MCV: 1789 (866 builds (88%), 53% wins)
APC: 1448 (362 builds (37%), 50% wins)
Light Tank: 1163 (180 builds (18%), 47% wins)
Ranger: 444 (271 builds (27%), 51% wins)
Mammoth Tank: 219 (30 builds (3%), 87% wins)
Tesla Tank: 219 (21 builds (2%), 48% wins)
Demo: 66 (27 builds (3%), 52% wins)
Phase Transport: 58 (23 builds (2%), 57% wins)
Chrono Tank: 45 (2 builds (0%), 100% wins)
Minelayer: 42 (24 builds (2%), 46% wins)
Radar Jammer: 22 (16 builds (2%), 75% wins)
Supply Truck: 10 (6 builds (1%), 33% wins)
Mobile Gap Generator: 2 (2 builds (0%), 50% wins)

SEASON 5
Medium Tank: 4257 (177 builds (37%), 59% wins)
Heavy Tank: 3425 (230 builds (48%), 50% wins)
Arti: 1313 (125 builds (26%), 56% wins)
Flak: 1290 (174 builds (36%), 45% wins)
Harvester: 1273 (445 builds (92%), 52% wins)
V2: 875 (146 builds (30%), 45% wins)
MCV: 869 (429 builds (89%), 51% wins)
Mammoth Tank: 304 (36 builds (7%), 72% wins)
Light Tank: 282 (88 builds (18%), 48% wins)
APC: 272 (137 builds (28%), 44% wins)
Tesla Tank: 230 (21 builds (4%), 76% wins)
Ranger: 171 (130 builds (27%), 60% wins)
Demo: 35 (19 builds (4%), 47% wins)
Chrono Tank: 33 (4 builds (1%), 75% wins)
Phase Transport: 17 (9 builds (2%), 67% wins)
Minelayer: 13 (12 builds (2%), 75% wins)
Radar Jammer: 12 (8 builds (2%), 75% wins)
Mobile Gap Generator: 7 (4 builds (1%), 75% wins)
Supply Truck: 5 (2 builds (0%), 0% wins)
Mad Tank: 5 (1 builds (0%), 0% wins)

SEASON 6
Heavy Tank: 4441 (314 builds (57%), 52% wins)
Medium Tank: 3888 (143 builds (26%), 57% wins)
Harvester: 1441 (501 builds (91%), 52% wins)
Flak: 1376 (207 builds (38%), 46% wins)
V2: 908 (165 builds (30%), 57% wins)
MCV: 905 (486 builds (88%), 52% wins)
Arti: 741 (77 builds (14%), 57% wins)
APC: 510 (192 builds (35%), 47% wins)
Light Tank: 470 (60 builds (11%), 46% wins)
Mammoth Tank: 435 (48 builds (9%), 75% wins)
Tesla Tank: 383 (28 builds (5%), 79% wins)
Ranger: 120 (99 builds (18%), 69% wins)
Minelayer: 25 (13 builds (2%), 31% wins)
Demo: 20 (18 builds (3%), 56% wins)
Supply Truck: 10 (5 builds (1%), 80% wins)
Radar Jammer: 7 (3 builds (1%), 67% wins)
Phase Transport: 3 (2 builds (0%), 50% wins)

SEASON 7
Medium Tank: 5982 (214 builds (36%), 54% wins)
Heavy Tank: 5233 (301 builds (51%), 51% wins)
Harvester: 1420 (538 builds (91%), 50% wins)
Flak: 1368 (252 builds (43%), 50% wins)
MCV: 1019 (549 builds (93%), 52% wins)
Mammoth Tank: 1001 (62 builds (10%), 73% wins)
Tesla Tank: 732 (34 builds (6%), 76% wins)
V2: 709 (148 builds (25%), 58% wins)
Arti: 630 (100 builds (17%), 54% wins)
Light Tank: 470 (90 builds (15%), 42% wins)
APC: 253 (158 builds (27%), 47% wins)
Ranger: 223 (174 builds (29%), 54% wins)
Radar Jammer: 33 (17 builds (3%), 94% wins)
Minelayer: 31 (22 builds (4%), 59% wins)
Demo: 22 (12 builds (2%), 17% wins)
Phase Transport: 16 (9 builds (2%), 67% wins)
Mobile Gap Generator: 9 (4 builds (1%), 100% wins)
Supply Truck: 3 (1 builds (0%), 100% wins)

SEASON 8
Medium Tank: 7507 (230 builds (40%), 53% wins)
Heavy Tank: 7163 (286 builds (49%), 53% wins)
Harvester: 1389 (531 builds (91%), 51% wins)
Flak: 1300 (241 builds (41%), 49% wins)
Mammoth Tank: 1233 (78 builds (13%), 71% wins)
MCV: 940 (534 builds (92%), 52% wins)
V2: 758 (163 builds (28%), 54% wins)
Arti: 688 (127 builds (22%), 44% wins)
Light Tank: 441 (118 builds (20%), 45% wins)
APC: 346 (135 builds (23%), 50% wins)
Tesla Tank: 344 (18 builds (3%), 50% wins)
Ranger: 312 (196 builds (34%), 53% wins)
Demo: 47 (31 builds (5%), 52% wins)
Radar Jammer: 42 (17 builds (3%), 59% wins)
Minelayer: 41 (22 builds (4%), 45% wins)
Chrono Tank: 9 (2 builds (0%), 50% wins)
Supply Truck: 6 (2 builds (0%), 50% wins)
Mobile Gap Generator: 4 (3 builds (1%), 33% wins)
Phase Transport: 3 (3 builds (1%), 33% wins)

SEASON 9
Heavy Tank: 16329 (421 builds (50%), 45% wins)
Medium Tank: 14131 (362 builds (43%), 58% wins)
Harvester: 2091 (789 builds (93%), 51% wins)
MCV: 1101 (786 builds (93%), 51% wins)
Flak: 1069 (304 builds (36%), 45% wins)
Mammoth Tank: 935 (67 builds (8%), 71% wins)
V2: 639 (158 builds (19%), 57% wins)
Arti: 555 (119 builds (14%), 60% wins)
Ranger: 523 (316 builds (37%), 57% wins)
Light Tank: 462 (142 builds (17%), 55% wins)
APC: 360 (190 builds (22%), 38% wins)
Tesla Tank: 344 (13 builds (2%), 77% wins)
Minelayer: 110 (89 builds (10%), 44% wins)
Chrono Tank: 52 (4 builds (0%), 25% wins)
Demo: 39 (23 builds (3%), 32% wins)
Radar Jammer: 21 (16 builds (2%), 62% wins)
Phase Transport: 14 (4 builds (0%), 75% wins)
Mobile Gap Generator: 2 (1 builds (0%), 0% wins)
Mad Tank: 1 (1 builds (0%), 100% wins)
Unsurprisingly the most queued vehicles are heavy tanks and mediums. In earlier seasons then artilleries top the table, which ties in with the greater popularity of radar play and Allies. I've heard from several sources that arti spam used to be the standard strategy in older seasons (although interestingly it only has a 49% S01 win rate compared with 58% for medium tanks).

The next graph takes the number of each unit queued and divides it by the number of times it was possible to queue it. There are twenty different vehicles (ignoring minor differences like the two types of minelayer) which is really too much for one graph, so I've tried to highlight the top four in colour.
019_vehiclesAllBuilt.png
019_vehiclesAllBuilt.png (49.48 KiB) Viewed 14843 times
After heavies we can see that flak tracks are the most popular unit for Soviet players. Given they're a common light vehicle pre-service depot and they're also useful post radar dome for harassing artilleries and anti-air then this is unsurprising.

In the next post then we'll have a look at medium tanks vs heavy tanks a bit more.

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

We saw in the previous post that medium and heavy tank production dominates all other vehicles. Here's the same graph again, showing that we're approaching the "OpenRA tank singularity" as SirCake put it:
020_vehiclesMTHT.png
020_vehiclesMTHT.png (39.2 KiB) Viewed 14840 times
Heavy tanks take longer to produce than mediums and so it would make sense for Soviet players to have shorter tank queues than Allies. However in Season 9 we saw the queue lengths roughly match. I think the reason the same number of heavies and mediums are now queued is predominantly due to the priority queue feature. An 'infinite' queue of tanks can be maintained and other items can be added to the front of the queue (using Ctrl). This was added in Release 20190314, which translates to RAGL Season 7. The explanation isn't particularly satisfactory though, since the medium-heavy gap actually widened between Seasons 6 and 7 (by an average of 0.03 tanks per game). Anyway we've continued to see tank queues increase in length in a similar way to rifle and rocket soldiers.

We mentioned earlier that building a service depot has increasingly become the only approach, and so it shouldn't be surprising to also see the production of medium and heavy tanks becoming the norm. Perhaps Soviet players have a few more rush options than Allies, and so the heavy tank line trails a bit below the medium tank line.
020_vehiclesMTHTBuildRate.png
020_vehiclesMTHTBuildRate.png (35.9 KiB) Viewed 14840 times
Finally I wanted to look at win rates for heavies and mediums. The win rates for MT and HTs are pretty much the same as the win rates for Allies and Soviets, but with the start of the graph shifted up a bit. Given how prevalent these tanks are then I also wanted to look at the win rates for players who didn't build any (but who did had the opportunity to).
020_vehiclesMTHTWinRate.png
020_vehiclesMTHTWinRate.png (50.43 KiB) Viewed 14840 times
Here we can see that skipping tanks is on the whole quite risky. For Allies it seems fairly clear that it's less than optimal. For Soviets then it didn't used to be good, but in Season 9 then it was comparable. Partly this seems to be because 'normal' builds by Soviets did worse than normal.

Also we can see that there's more error in the results for skipping tanks, because fewer players did it. In fact there were so few replays where Allied players skipped medium tanks that I thought it was worth having a look through them. From Season 9 we just had the following list of games where an Allied player won without a medium tank:

Code: Select all

SinJul RAGL-S09-MINION-GROUP-GOR-SIN-G2.orarep Ranger Danger
WhoCares RAGL-S09-MASTER-GROUP-GOA-WCS-G2.orarep Triple-Ref-All-In
WhoCares RAGL-S09-MASTER-GROUP-GOA-WCS-G1.orarep Triple-Ref-All-In
morkel RAGL-S09-MASTER-GROUP-MRK-FIV-G1.orarep 3:35 gg
Kernel Panic RAGL-S09-MINION-GROUP-S84-KNL-G2.orarep Ranger Danger
SinJul RAGL-S09-MINION-GROUP-S84-SIN-G2.orarep Defended against Double-Ref-All-In
Zaqzorn RAGL-S09-MINION-GROUP-ZAQ-KNL-G2.orarep Double-Ref-All-In
The matches between Goat and WhoCares are particularly worth watching, as they are very high quality examples of how to Triple-Ref-All-In.
Last edited by TTTPPP on Thu Sep 03, 2020 12:16 pm, edited 1 time in total.

Yara
Posts: 28
Joined: Tue Feb 19, 2019 7:22 am

Re: RAGL through the seasons

Post by Yara »

Do you have also access to the chat logs? Would be fun to see who wrote "gg" at the end and who did not ^^

TTTPPP
Posts: 320
Joined: Wed Sep 05, 2018 8:25 pm

Re: RAGL through the seasons

Post by TTTPPP »

Yep - the chat messages are available from the games. I did a bit of investigation into common messages in an earlier post: viewtopic.php?p=312071&sid=c4b05a909d16 ... bd#p312071 In particular I looked at the correlation between saying "gg" first and losing the game. I'm happy to look into something else in this area if you have a suggestion, although grouping games by player is quite hard since most seasons were before it was possible to log in.

User avatar
Punsho
Posts: 143
Joined: Wed Jul 18, 2018 2:56 pm
Location: Lithuania

Re: RAGL through the seasons

Post by Punsho »

Perhaps less heavy tanks are queued because they take longer to build -> you build less of them

Post Reply