Winning streaks and losing streaks

[kirkmc]

Sometimes I just don't understand why I play the way I play. A month or so ago, I had what's probably the longest winning streak I ever head: I think i won 18 games in a row. For almost all of these games, all the right moves just came naturally (well, not all, but enough to win). It seemed easy to win, and I shot up from high 5K to low 4K pretty quickly. Then I settled in at 4K, and, a few weeks later, started losing. Not every game, but in July I lost more than 60% of my games, and had days when every move I played was wrong. Now I've won my last six games in a row.

This happens, I know, but it's never happened to me in such stark contrast. I'm curious as to what can cause this. I know it's not a question of skills: even though during my losing streak I felt I simply didn't know what moves to play, the knowledge is still there. But it just seems that I couldn't bring that knowledge forward and access it.

Any thoughts on this?

[Mike Novack]

Sometimes I just don't understand why I play the way I play. A month or so ago, I had what's probably the longest winning streak I ever head: I think i won 18 games in a row. For almost all of these games, all the right moves just came naturally (well, not all, but enough to win). It seemed easy to win, and I shot up from high 5K to low 4K pretty quickly. Then I settled in at 4K, and, a few weeks later, started losing. Not every game, but in July I lost more than 60% of my games, and had days when every move I played was wrong. Now I've won my last six games in a row.

This happens, I know, but it's never happened to me in such stark contrast. I'm curious as to what can cause this. I know it's not a question of skills: even though during my losing streak I felt I simply didn't know what moves to play, the knowledge is still there. But it just seems that I couldn't bring that knowledge forward and access it.

Any thoughts on this?

Probably artifact of not understanding probability and looking at this from the perspective of yourself. People tend to grossly underestimate the expected "streakiness" of a true random distribution. And though it might be very unlikely that you win a lottery there is somebody who wins it. For example, if we have 1000 people playing go on a regular basis it is very likely that some of them win or lose eight in a row. If you like, take eight coins, toss them a thousand times, and note down how many times came up all heads or all tails.

[Monadology]

But probability doesn't explain the phenomenon that was described as causing the problem, which is the inability to make correct plays where one recently was perfectly capable of doing so.

I've had the same experience but without winning or losing streaks since it doesn't seem to last much more than a day.

[gowan]

There could be several factors involved in this phenomenon. Even though Go is a cerebral, intellectual pursuit your play often depends on your mood or emotional state. Winning induces a euphoric state which itself induces more winning, hence the winning streak. When you reach a higher rank, that may be accompanied by some anxiety about being able to perform at that level so you lose. This then induces a down emotional state which leads to more losses. When you are in a down emotional state you might well not be able to muster your full go knowledge. Of course I don't know whether this explains your situation.

[palapiku]

You were underranked at 5k, got paired against weaker players, crushed them until your rank fixed itself, now you're back to 50% winning probability.

[karaklis]

You have more than 100 matches per month. Winning/losing streaks of 6-7 consecutive matches should be pretty normal with such an amount of matches.

[kirkmc]

You have more than 100 matches per month. Winning/losing streaks of 6-7 consecutive matches should be pretty normal with such an amount of matches.

6 or 7 yes, but I'm talking about 15-20, which is something I've never had before.

[hyperpape]

Sounds like we need a statistician. Given the right data set, we could test whether go victories seem to be streaky or whether the results of consecutive games are relatively independent. Surely the answer is somewhere in the middle, but exactly where is an interesting question.

[kirkmc]

Sounds like we need a statistician. Given the right data set, we could test whether go victories seem to be streaky or whether the results of consecutive games are relatively independent. Surely the answer is somewhere in the middle, but exactly where is an interesting question.

Small streaks are certainly common; a half-dozen games one way or another is something I see often. But rarely more than ten in a row, unless, of course, someone is a sandbagger, and even then their rank goes up quickly enough in such cases that streaks like that don't last.

[Wildclaw]

Then I settled in at 4K, and, a few weeks later, started losing. Not every game, but in July I lost more than 60% of my games,

Match ups using a traditional handicap game system are far from 50-50.

• If you are a weak (strong) player in any rank you will on average have half a stone disadvantage (advantage).
  
• Also, when playing white (black) in a handicap game you by default have an additional half stone advantage (disadvantage) due to the lack of komi

The estimated chance of winning depends mainly on the numbers the specific system uses to create then scale. Using the KGS figures (if they are still accurate?), you will have this kind of chance of winning depending on the advantage. The first number is for 5kyu or weaker, the second for 2dan or stronger players. Example match ups within parenthesis.

• Even - 50-50% (a weak 4 kyu or average 5 kyu vs a weak 4 kyu)
• Half a stone - 60-66% (an average 4 kyu or strong 5 kyu vs a weak 4 kyu)
• One stone - 70-79% (a strong 4 kyu or average 3 kyu vs a weak 4 kyu)
• One and a half stone - 78-88% (a strong 3 kyu vs a weak 4 kyu)

Having an 18 game winning streak is pretty rare, but not extremely rare. With a half stone advantage on average as a strong player in a rank you will have to play about 25000 games to experience that. Not a miracle exactly. And that is before even considering sudden changes in rank (so you are more than half a stone stronger on average), mood, general barriers, etc.

Having a 60% loss rate as a weak 4kyu on the other hand is just plain average.

[kirkmc]

Having a 60% loss rate as a weak 4kyu on the other hand is just plain average.

I don't think that's correct. Having a 60% loss rate would lead to a lower rank. 50%, if I understand the KGS rating system, is average. (That is, 50% against players of your strength. Obviously we'll always be playing against stronger and weaker players. But the handicap system should lead to a 50% average over time.)

[flOvermind]

Yes, if you're 4.00k, you'll win 50% against other 4k players (which I assume are evenly distributed from 3.5k to 4.5k). But if you're, for example, a 4.45k, that's almost half a stone away from the average 4k. In that case, 60% does sound about right.

Of course that's all just statistics, and only meaningful with a large sample. I realize that individual ranks are not anywhere accurate enough to make two fractional digits meaningful. But it still shows that each winning percentage from about 40% to 60% is realistic for even games within a single rank. A 60% loss ratio just indicates that you are slightly on the weak side of 4k. But you're still nearer to 4k than to 5k, because for a whole rank difference, I would expect more like 70-75%.

As for the unusually long winning streak: Perhaps you really improved, and learning something that the average KGS 5k doesn't know but the average 4k does. Such things can produce pretty scary streaks until you find an opponent who can deal with it. Even if you don't know anything specific, such things can happen subconsciously

[kirkmc]

As for the unusually long winning streak: Perhaps you really improved, and learning something that the average KGS 5k doesn't know but the average 4k does. Such things can produce pretty scary streaks until you find an opponent who can deal with it. Even if you don't know anything specific, such things can happen subconsciously

That's what I thought. But how come I've forgotten whatever it was I learned already?

[Mike Novack]

Having an 18 game winning streak is pretty rare, but not extremely rare. With a half stone advantage on average as a strong player in a rank you will have to play about 25000 games to experience that. Not a miracle exactly. And that is before even considering sudden changes in rank (so you are more than half a stone stronger on average), mood, general barriers, etc.

And more to the point perhaps. Playing 25,000 games sounds like a lot but a lot of people playing go. So for a group of 1000 players each about 1/2 stone strong for their rank and each playing 30-40 games pretty likely that at least one of them experiences a winning streak that long.

Try looking at this a different way. Imagine there are 256 people of equal playing strength entered into a "knock-out" tournament. It might be unlikely for any particular player to win but one of them must win eight in a row (that's certain).

[flOvermind]

As for the unusually long winning streak: Perhaps you really improved, and learning something that the average KGS 5k doesn't know but the average 4k does. Such things can produce pretty scary streaks until you find an opponent who can deal with it. Even if you don't know anything specific, such things can happen subconsciously

That's what I thought. But how come I've forgotten whatever it was I learned already?

You are ranked 4k now, aren't you? Assuming you can hold that rank and don't fall back to 5k, it may be that you haven't forgotten, but are just playing more opponents who know it, too

(Of course that argument has the little flaw that we don't know what "it" is. Never let the facts get in your way )