World of warships losing streaks

I wrote about losing streaks earlier, but I recently realized a methodological error. I treated the games as a stream, one game after the other, while games take place in sessions. If I play a game in the evening, go to sleep and play the next one in the morning, it’s not really the “next” one. So my early data is noisy. It’s not useless, because a game is either followed by the next or a random game, so the signal is still there and there is no anti-signal. However at the last 280 games I properly recorded session limits.

sessions

There is an agreement between the two methods in that if we look at the previous game, it predicts the next one. It gives 54-57% winrate after a won game and a 43-47% winrate after a lost game. They also agree that Win-Win predicts Win and Lose-Lose predicts lose. They disagree with the Win-Lose prediction though. Finally, the proper method says that the first game of the session has the same winrate as the games after wins. Ergo, there is no downside interrupting a “winning streak”, it seems there are no winning streaks, just losing ones. The win-chance after a win is just the win chance.

From this I take away a very simple rule: one hour off after a defeat. It helps if the problem was me (tired, distracted), if the problem was the immediate team (hating) or if the problem is the meta (too many destroyer players for example). As the above result shows, there is no downside of taking a break.

Let’s see the intersession correlations. There were 60 sessions recorded with 280 games. Their outcomes are:

  • 19 times a loss session was followed by loss
  • 7 times a win session was followed by a win
  • 9 times a loss session was followed by a win
  • 7 times a win session was followed by a loss
  • Sessions could be neutral (like WWLL) and those aren’t counted
    • .
      Again we see clearly not random amount of loss-streaks. These mean longer, whole-day trends, so as an auxiliary rule, if two sessions ends in negative outcome, a much longer break is needed.

      I’m sure that such stop-loss methods will greatly increase my winrate.

Author: Gevlon

My blog: https://greedygoblinblog.wordpress.com/

17 thoughts on “World of warships losing streaks”

  1. You are treating your win rate as constant, which it clearly wasn’t. You played the first 150 or so games at 55%, then 400 games at 50%, then last 100 at 57%. All the games below 2-5 bracket should be discarded.
    If you have the records of date and time of the games you will show something meaningful, maybe. Predicting win/loss based on the previous game is silly when the outcome is so random and there are so few sessions. Your “first game” results are 32-28. This means that every single game influenced the overall result by 1.7%. The error margin is larger than your signal.

    Like

  2. I’m sure this just noise. You try to fund patterns – thus you find them.

    The issue with predominately random created outcomes of just two cases (win loss) is that almost anything looks like a pattern.

    Think about all the myths of uncontrolled coin flips – e.g. child sex during birth. Girl gets ugly = “it’s a girl!”. Has to vomit “it’s a boy”

    You just randomly hit too often to make these assumptions. Especially “it doesn’t matter whats the root cause, stopping to play matters.”
    Is the worst overreaction you can make and you try to control the uncontrollable. Think about poker: win or loss in a single round is random. If you realize loosing too much at one table (= session) you are either too bad against the available playerpool or your strat doesn’t work against their gamestyle. Most important is: dont play on tilt. Dont play to catch up, dont play after a bad beat. Dont play tired.

    THAT’S what you control – nothing else. Predicting WL outcome just with one game in advance is statistical nonsense.

    Like

  3. @Stawek: 200 vs 153 and 153 vs 176 is pretty significant. So even if we discard everything else, it stands.

    @Chewiecide: even if you have nothing but “don’t play on tilt”, how do you measure “tilt”? The only objective measurement we have is “tilt causes losing”. So all we have left is “stop playing if you’re losing”.

    Like

  4. You control your tilt. Its called anger management. Thats the most important skill in any competitive PvP environment.

    So either you tilt directly after a loss which makes your skill ceiling in pvp games pretty Limited or you don’t have a clue why you are losing which makes your strategic wows capabilities pretty limited.

    Just think for a second: “just pick shima and spam torps and cap if safe – if you lost a single game pause an Hour“ this is your gold guide for average joe to Rank out? Breaking the Meta? The more you blog on wows the more I realize that this is Really a Game where all your efforts will fail. Simply due to the fact that it’s a team game with non ELO matchmaking and your impact of Outcome hardly exceeds 15%.

    A tip: recruit 7-10 players who have enough time to grind CW season and try to prove that basic strat in all ships allowed without voice is enough to become a purple clan. That would break the meta. Not your playstyle in ranked which by definition can’t have a big impact on outcome.

    Like

  5. @Gevlon

    “So all we have left is “stop playing if you’re losing”

    You are claiming(in another post) that the player pool is causing this, though. If the player assignment is completely random, then the only thing left to determine win rate is the skill of the player. That is, unless you are hinting that the match maker is taking certain variables into account that it shouldn’t be.

    Like

  6. @Chewiecide: we’ll see. If next season I rank out in 100-200 games, we have an evidence. I do believe that most games can be broken by ridiculous meta’s. Like PUBG could be broken by “hide in the bathroom”.

    @NoGuff: I’m GUESSING that the player pool is causing it. I don’t really care. I look for “objective input -> objective action” rules.

    Like

  7. @Gevlon

    ” I’m GUESSING that the player pool is causing it. I don’t really care. I look for “objective input -> objective action” rules.”

    So how does that play into the equation when Stawek has indicated multiple times that it is more difficult to rank out “early”? If the higher skilled players rank out faster, then the skill level of the remaining players(opponents) has to drop over time also. So you being able to rank out in 100-200 games will be entirely a function of your actual skill, versus how early or late in the season that you start trying to rank out.

    Like

  8. Gevlon, I still think you are kinda data dredging here. Sure, there might be slight correlation between how you feel and how you play and therefore making a stop, if you play shitty today, is a good idea. But overall, you put too much faith into these small data sets. If you get more and the trend continues, then sure, right now it’s feels somewhat premature.

    Like

  9. even if you rank out early next season you have proven nothing besides that you learned the game after a couple of thousand games. Now what? You did not follow a new strategy nor did you break any meta. Its as ordinary as it can get and you did nothing special than – just playing the game? The effort of ranking out in <200 games is skill based. Not a mysterious strategy that can be copied.

    Like

  10. @Chewiecide: the real test is.
    – I rank out early with good results
    – I write a strategy guide
    – people who copy it also rank out, despite their previous stats didn’t predict that

    Like

  11. But who should follow it? Lets assume you need 200 games to rank out. That’s 50 hours. Than your stop when losing rule kicks in. Lets assume you lose every third game. And pause for half an hour. Thats another 34 hours. Thats 2 hours / day attached to the game for 6 weeks. Spaming torps in a shima.and leaving for 30 min breaks where even top 1% cant get more than 66% winrate?

    Who on earth would follow that guide? I predict something: you will claim that you have been right. Create a stable page and next game.

    Like

  12. It will be a meta-guide, not strategy. I can write it for you: wait until the last 2 weeks of the season and don’t play peak times.

    There. It won’t make people play any better but they will feel like pros getting R1 in 200 games and not 400.

    Like

  13. @Chewiecide: everyone who wishes to rank out already devote 50+ hours to the game, so it’s a non-issue. The waiting periods aren’t added, because you can do other things.

    @Stawek: no, it will be both a meta and a strategy guide (how to fit and use the Shima)

    Like

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s