Red Alert Global League: Season 13 - Standings and Statistics

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

At this point in the RAGL Through The Seasons thread I looked into common messages sent by players. Since I already had the code written I thought I might as well extend the stats up to Season 13.
013_RallyingCryPopularity.png
013_RallyingCryPopularity.png (33.13 KiB) Viewed 6534 times
The first thing that stands out here is that these numbers don't match the ones I got previously. For example - last time I thought 50% of players said "glhf" in Season 8, but this time it's nearer 65%. Obviously I've made a mistake in one of these calculations (or maybe both?) but I've checked the query this time and it looks right, so I'm inclined to think these are the real numbers.
013_RallyingCryWinRate.png
013_RallyingCryWinRate.png (41.18 KiB) Viewed 6534 times
Checking out the win rates then we find that "gl" is the best message to send in the modern game. Incidentally I tried various typo versions and you can pretty much find a game for any key slip that you like - something to bare in mind if you're worried that other players might be perfect.
013_GGWinRate.png
013_GGWinRate.png (23.28 KiB) Viewed 6534 times
Finally here's the win rate of players who said "gg" (or "gg wp") before their opponent. We can see that it's generally fallen over time, but might have started picking up again in Season 13.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

014_GameDuration.png
014_GameDuration.png (31.31 KiB) Viewed 6504 times
The first few seasons had some crazily long games. Season 13 featured the longest game since Season 4 (relive that pain with Kaution's cast here). The average game length has stayed pretty constant though.
014_GameLengthDistribution.png
014_GameLengthDistribution.png (41.42 KiB) Viewed 6504 times
Only 10% of games are longer than 20 minutes, the vast majority are between 5 and 20 minutes.
014_FactionWinRates.png
014_FactionWinRates.png (39.74 KiB) Viewed 6504 times
Finally I revisited faction win rates after considering that usually I include matches where both players have a particular attribute. This graph shows the kind difference it can make when we look at the win rate of e.g. Allies against "not Allies". The "vs Anyone" lines are a repeat of what we saw in an earlier graph.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

In this post I wanted to look at a few factors that might impact game length.
015_AverageGameDurationsByDivision.png
015_AverageGameDurationsByDivision.png (35.98 KiB) Viewed 6488 times
First we can see that Minions have recently had longer games than Masters, but in Seasons 9, 10 and 11 there wasn't much difference, and before this Masters had longer games. Potentially this is due to a low skill gap in Minions, or maybe it's players failing to close out games.
015_AverageGameLengthByFactions.png
015_AverageGameLengthByFactions.png (49.6 KiB) Viewed 6488 times
Looking at how faction impacts game length then there's no obvious pattern. I originally hoped to find evidence of stalemates in Allies late games, but even if it's here then it's not significant enough to show through.
015_AverageGameLengthByBuildOrders.png
015_AverageGameLengthByBuildOrders.png (86.61 KiB) Viewed 6488 times
Finally I used the build order list from earlier to compare game lengths. Comparing this with the win rate graph from earlier then we can see the shorter games are for the 'worse' build orders ("Other PP,Rx,Rf,Rf" and "Other"), but the longer games don't necessarily correspond to better build orders. In Seasons 12 and 13 one of the shortest game lengths was for "PP,Rx,Rf,Rf,PP,WF,Rf,Rx" which was one of the best performing build orders, so there's another exception to the rule.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

016_SpyUsageByFaction.png
016_SpyUsageByFaction.png (47.85 KiB) Viewed 6447 times
One of the faction bonuses is the English spy. From this graph it's not immediately clear whether English players build spies more often or not, but the average across all seasons says that England builds spies about 11% of the time, whereas France and Germany are both at 8%. Weirdly in Season 6 there were no England players who built any spies.
016_SpyCountByFaction.png
016_SpyCountByFaction.png (48.9 KiB) Viewed 6447 times
When a player builds spies then they're likely to build slightly more with England too. This time the overall average figures are 3.8 spies per England player (building spies), compared with 2.9 for France and 3.5 for Germany. There are some pretty huge spikes in this graph, which suggest that there's probably one or two players/games biasing the overall results.
016_SpyWinRateByFaction.png
016_SpyWinRateByFaction.png (41.6 KiB) Viewed 6447 times
Finally if we look at win rates for players who built spies we can see that the data is again all over the place. The average across all seasons says English spies win 47% of their games, whereas French and German spies are at 65% and 69% respectively. I find this result hard to explain especially because the overall England win rate is very slightly higher than the other two subfactions (53% vs 52% and 51%). Potentially England players are being tempted into building spies when they normally wouldn't? Anyway it seems fair to say that the discount for the English spy isn't a bonus that many players are converting into victories.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

017_RifleRocketRatioThroughQueue.png
017_RifleRocketRatioThroughQueue.png (65.21 KiB) Viewed 6395 times
We looked at the rifle-rocket ratio earlier on, but we didn't look at how it develops through the average game. The scale at the bottom of the graph is a bit hard to read, but each notch represents a batch of 50 infantry. We can see from this that usually the rifle-rocket ratio is about 6:1 for the first 50 infantry, and then drops from 3:1 towards 2:1 as the game goes on. This pattern has been present in pretty much all seasons.
017_RifleRocketCountsByDivision.png
017_RifleRocketCountsByDivision.png (34.9 KiB) Viewed 6395 times
I had a request to investigate rifle-rocket ratios for different players, and this was hard to present the data in a particularly meaningful way. However I was able to plot the number of rifles and rockets queued for every game in each division. Masters and Minions look pretty similar, but Recruits generally seemed to be queueing fewer units.

The raw data for individual player's rifle-rocket ratio across all seasons is below (e.g. summing across all games then Follicle has queued roughly 10 rifles for every rocket):

Code: Select all

Follicle: 10.2
porcukor: 9.2
DVoid: 7.1
Иαιgεl: 7.0
Snufflegrunt: 6.8
Julle: 6.8
toiletbreakbrb: 6.7
Insomnic: 6.5
JuiceBox: 5.9
Chin0: 5.7
KOYK_GR: 5.7
Spy: 5.7
ConcaveCats: 5.6
kyrylo: 5.5
Chaosgrunt22: 5.5
potato: 5.4
merimursu: 5.4
AmericanBlunt_: 5.1
ZxGanon: 5.1
jungle: 4.9
worldpeace: 4.8
bobbyd: 4.6
Abarrat: 4.5
Strateeg: 4.5
Chewey: 4.4
DomoTheRussian: 4.3
raymundo: 4.3
OMnom: 4.3
J!NX: 4.3
Fahrrad: 4.1
TiTo: 4.1
JackoDerp: 4.1
ROCKhardFISTnips: 4.1
beuk: 4.0
Azure Anemone: 4.0
Daz: 4.0
BigSmoke: 4.0
Fundahl: 4.0
WaRRyKeR: 4.0
Oafa: 4.0
MicroBit: 3.9
h34vy: 3.9
mechANIC: 3.9
Declan: 3.9
vvoj: 3.9
fittsaft: 3.9
PieChild: 3.8
Murto the Ray: 3.8
squidgeny: 3.8
Mo: 3.7
MrZuerg: 3.7
jvk: 3.7
DrLuke: 3.7
anjew: 3.6
Yara: 3.6
Beef: 3.6
abcdefg30: 3.6
Mentats: 3.6
eskimo: 3.5
hamb: 3.5
Mr Cloudy: 3.5
MR C: 3.5
Orb: 3.5
crash_dash: 3.5
Metalcallous: 3.5
Clockwork: 3.5
Margot Honecker: 3.4
copla201: 3.4
PallyRoy: 3.3
shavkat: 3.3
RobLoweEscobar: 3.3
Aggr3ss1vePanda: 3.3
KSK_Nico: 3.3
Misery: 3.3
Kuba: 3.3
fris: 3.3
chouchani: 3.2
jjkramok: 3.2
dragunoff: 3.2
Goremented: 3.2
Balloon: 3.2
NethIafin: 3.2
Sigil: 3.2
weeman: 3.2
Nukin': 3.2
BlueJay: 3.2
Icesinithol: 3.2
Dualwarhead: 3.1
Upps: 3.1
Karthus: 3.1
_Bo: 3.1
DodoBirb: 3.1
Pvt_Leaf: 3.1
malboo: 3.1
Zaqzorn: 3.1
Petit Dejeuner: 3.1
poop: 3.1
netnazgul: 3.1
morkel: 3.1
jreed: 3.1
Submarine: 3.0
Blackened: 3.0
3.Lucian: 3.0
FiveAces: 3.0
SoScared: 3.0
goat: 3.0
Fruttielicious: 3.0
Craig Christ: 3.0
Han: 3.0
ioverthoughtthis: 3.0
i like men: 2.9
Doomsday: 2.9
WhoCares: 2.9
Sotsch: 2.9
Mint Spokesman: 2.9
Pinkthoth: 2.8
Dodder: 2.8
Wippie: 2.8
tux: 2.8
BigHALK: 2.8
spetsnaz84: 2.7
bullE: 2.7
Punsho: 2.7
klaas: 2.7
Raishiwi: 2.7
lundiz: 2.7
bete: 2.7
barf_openra: 2.7
LaKulta: 2.7
SinJul: 2.7
Gajcus: 2.7
SlickSloth: 2.7
maceman: 2.7
kazu.: 2.6
Ikke: 2.6
corrode: 2.6
Kernel Panic: 2.6
Kav: 2.6
SarahSicaria: 2.6
Printer: 2.6
Anykeyich: 2.6
Ohmk Torque: 2.6
Medium Tank: 2.5
despro: 2.5
Strelkov: 2.5
Therapist: 2.5
jaymanxyz: 2.5
OzzyOuzo: 2.5
Testosterone Rex: 2.5
milkman: 2.5
Duke Bones: 2.5
t: 2.5
TTTPPP: 2.4
Tailix Killa Mentor: 2.4
Lord of War: 2.4
Dying Fetus: 2.4
Nilhall: 2.4
Eugenator: 2.4
ass_pounder_4000: 2.4
zoidyberg: 2.4
lucassss: 2.4
bµg: 2.3
dang_shot: 2.3
Mees: 2.3
Jur: 2.3
Smitty: 2.3
Barxx: 2.3
noobmapmaker: 2.3
Karina: 2.3
chrisf: 2.3
Lorrydriver: 2.2
Gatekeeper: 2.2
Jamesasthenia: 2.2
ThunderStruck115: 2.2
manta151: 2.2
Ekanim: 2.2
SirCake: 2.2
Hermann G': 2.1
SeaCucumber: 2.1
Henry_the_Slav: 2.1
Bamsi: 2.1
Flecken: 2.1
Kallen: 2.0
Fenix: 2.0
Polaris: 2.0
FRenzy: 2.0
Digital Jihad: 2.0
xy: 2.0
Odourless: 2.0
Kaution: 2.0
daswaseinbefel: 1.9
flamewh33l: 1.9
shpooky: 1.9
Snake: 1.7
Antares: 1.6
Aaron_Lloyd: 1.6
DaolDnaKcol: 1.4
GoldenHippo: 1.4
manu200: 1.3
DoDoCat: 1.3
CottonTheButton: 1.3
bone|superr0x: 1.1
General: 1.1
FasTSlacK: 1.0
KillaB: 0.6
Greasyudders: 0.4
017_InfantryQueueComposition.png
017_InfantryQueueComposition.png (64.3 KiB) Viewed 6395 times
Finally a crazy graph showing how each block of 50 infantry is made up. For example the infantry between unit 351 and 400 are about 1% flamers. No more comments from me about this, but it was fun to look at.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

018_PercentageOfQueueConsistingOfGrenadiers.png
018_PercentageOfQueueConsistingOfGrenadiers.png (66.62 KiB) Viewed 6328 times
The overall popularity of grens has varied from season to season, but the general pattern is pretty much the same. Grenadiers are usually among the first ten units built, occasionally coming a bit later, but by the mid-game they are pretty much non-existent. This makes sense, since grens explode when they die which makes them a bad unit to mix into your army, and early game often gives you some easy unprotected routes to your opponent's base.
018_GrenadierCountPopularity.png
018_GrenadierCountPopularity.png (53.73 KiB) Viewed 6328 times
Now we know when the grens are coming, but how many should we expect? This graph shows that 4-6 grens was overwhelmingly the most popular option in Season 13. Other recent seasons also saw 2-3 grens as a common choice, but it's uncommon to see a single grenadier or more than seven.
018_Relationshipbetweengrenadiersanddogs.png
018_Relationshipbetweengrenadiersanddogs.png (52.06 KiB) Viewed 6328 times
Sarah raised an interesting possibility for changing the Soviet faction bonuses - give both subfactions shock troopers and maybe parabombs in exchange for grens being only Ukraine and dogs being Russian. The obvious downside is that players will no longer be able to use dogs and grens together, and this graph aims to look at how significant that would be. Averaging across all seasons then we see 58% of Soviet builds have neither unit in them, 19% have grens, 17% have dogs and 6% have both. If the two events (building grens and building dogs) were unrelated then you'd expect roughly 3% of builds to have both, so there's a significant increase over this - i.e. there's a correlation between a player building a dog and a player building a gren. It therefore seems a shame to remove this option, but at the same time it only affects 6% of Soviet builds, so it impacts quite a small number of games. Additionally by moving the other bonuses to both factions there are new interactions available.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

019_ShockTrooperUsage.png
019_ShockTrooperUsage.png (55.39 KiB) Viewed 6294 times
Shock troopers have a number of barriers to use them. Firstly you need to be Russia (or obtain Russian tech), something that only happens about 25% of the time for the many players picking "Any." Secondly you need to own a tech center and a teslacoil (and also a barracks obviously). This graph shows us that there's a significant percentage of players who achieve these goals and still don't add shock troopers to their queue. Season 13 represents a recent high in shock trooper usage with about 17% of Russian players building them. Seasons 10 and 11 in particular have very few shockies and so these data points are going to be a little patchy below.

Incidentally shockies have only been queued twice by non-Russia players - once in Season 4 by Craig Christ, and once in Season 9 by FiveAces.
019_ShockTrooperWinRate.png
019_ShockTrooperWinRate.png (39.61 KiB) Viewed 6294 times
As we noticed above Seasons 10 and 11 have fairly meaningless data points, but excluding these we see shock troopers giving at best a tiny advantage over Soviet T3 in general.
019_ShockTroopersAsPercentageOfRussianQueues.png
019_ShockTroopersAsPercentageOfRussianQueues.png (58.36 KiB) Viewed 6294 times
Finally I wanted to look at the proportion of the queue that's made up of shock troopers. There seems to be a steady decline in shock trooper spam over the seasons. Shockies only make up 3% of a typical Russian player's queue by the 400 unit mark.

I wanted to try to get an estimate of the ratios being used for shock troopers once the prerequisite buildings were built, and so the final line on the graph only includes Russian players in Seasons 11-13 who built a tech center at some point. The end of this graph is at about 13% and the equivalents for rifles and rockets are 60% and 24% - suggesting maybe a 5:2:1 ratio of rifles to rockets to shockies. We know that some Russian players don't build shock troopers even if they have the option, so maybe it's an underestimate of the number of shock troopers.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

020_MainTanksQueuedPerGame.png
020_MainTanksQueuedPerGame.png (38.15 KiB) Viewed 6238 times
On the whole the average number of queued tanks has risen over history of RAGL. There was a time when queuing ten tanks was enough, but nowadays the greater frequency of service depots has led to tanks being built from near the start of almost every game. We'll see from the next graph that this explanation doesn't explain why the averages have continued to rise since Season 10 though.
020_TankUsage.png
020_TankUsage.png (31.4 KiB) Viewed 6238 times
The proportion of builds including these tanks has also risen and peaked in Season 10 at an incredible 99.6% of Allied builds. The high for Soviets was in Season 12 with 97.6% of builds containing a heavy tank.
020_WinRateForMainTanks.png
020_WinRateForMainTanks.png (54.18 KiB) Viewed 6238 times
Looking at the win rates then we can see medium and heavy tank usage heading towards 50% as more and more builds contain them. The deviation around Seasons 9 to 11 can be explained by the overall win rate of Allies being higher than Soviets. There aren't as many builds that skip tanks entirely, but I thought it was interesting to have a look. For the most part we see that skipping tanks has been a risky choice, but in Season 13 the 10% of builds that skipped heavy tanks had a 64% win rate.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

021_LightVehiclesQueued.png
021_LightVehiclesQueued.png (54.29 KiB) Viewed 6122 times
The average number of flaks built by Soviet players has roughly followed the popularity of the radar dome. The ranger has roughly stayed at an average of one per Allied build, with a small peak in Seasons 8 and 9 which I think corresponds to Ranger Danger builds being more popular. Light tanks and APCs have followed an interesting trend - with more light tanks spammed on average in Season 13 than any previous season.
021_LightVehiclePopularity.png
021_LightVehiclePopularity.png (48.6 KiB) Viewed 6122 times
Looking at the popularity of light vehicles then we can confirm the increase in light tanks is due to players queuing more of them (rather than more players queuing them). Season 13 was the first in a while when the ranger wasn't the most popular Allied choice.
021_LightVehicleWinRate.png
021_LightVehicleWinRate.png (56.77 KiB) Viewed 6122 times
For win rates then I've included the win rate of not building each option. All of these lines are represented by a significant number of games, and most of them track pretty close to 50% win rates. It does seem that not building a light tank might give a small advantage though. At the other end we can see not building a ranger or a flak were poor performing options in Season 13.

The poor performance of the light tank might be due to players queuing them after their service depot was destroyed, and then going on to lose the game. We'll investigate this more next time by restricting our data to the first five vehicles.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

022_LightVehiclesQueuedInFirstFive.png
022_LightVehiclesQueuedInFirstFive.png (54.69 KiB) Viewed 6110 times
If we restrict the view to just the first five vehicles queued then we get a better idea of what vehicle is being used for early pressure. This does exclude light tank spam but gives a more even comparison of the four light vehicle options. We can see that in Season 13 they were incredibly even, and since they were all slightly above half a vehicle per queue then we can also tell there was a reasonable number of double light vehicle builds in here.
022_LightVehiclesUsageInFirstFive.png
022_LightVehiclesUsageInFirstFive.png (50.85 KiB) Viewed 6110 times
At the start of the game then the ranger is still the most popular option. Players are able to make use of it to get a vision advantage in engagements, ship their engi around the map and scout for rushes.
022_LightVehicleWinRateFirstFive.png
022_LightVehicleWinRateFirstFive.png (63.24 KiB) Viewed 6110 times
In terms of win rate then ranger and flak are the better options. There's not much in this though and we can see the flak's performance has swung wildly over the seasons.

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

Re: Red Alert Global League: Season 13 - Standings and Statistics

Post by TTTPPP »

023_HarvesterAndMCVCount.png
023_HarvesterAndMCVCount.png (30.21 KiB) Viewed 6064 times
This graph shows the average number of harvesters and MCVs queued - i.e. it excludes the harvesters that come with refineries and the MCV that you start with. This has been pretty consistent for all 13 seasons, with a small peak in the MCV line when using three MCVs was a fairly common strategy.
023_HarvesterAndMCVUsage.png
023_HarvesterAndMCVUsage.png (27.47 KiB) Viewed 6064 times
Most builds include at least one MCV and several harvesters, but even in Season 13 there was a healthy chunk of cheese builds including neither.
023_HarvesterAndMCVWinRate.png
023_HarvesterAndMCVWinRate.png (44.53 KiB) Viewed 6064 times
In terms of win rates then the ubiquity of these vehicles means that they're both at 50%. In the occasions when a player didn't build them then it generally paid off less than 50% of the time. There have been a few seasons where forgoing a harvester or MCV made it to a 50% win rate though. Interestingly the choice of no MCVs or harvesters was only successful in Season 1.

Post Reply