World Championship Predictions

[Barry Sandercock]

Magnus Carlsen will win. ( He always does.)

[Matt Mackenzie]

Well, usually

[Alistair Campbell]

Carlsen has a rating advantage of 81 points, which is an expected score of 0.61 per game. Not sure what the underlying distribution of results would be, though.

You could model the match as a series of 12 independent trials with probabilities of results as

0.3 Carlsen win
0.6 draw
0.1 Carlsen loss

(which gives an expected Carlsen score of 0.6 per game) - I think this is a trinomial distribution.

I calculate that this gives a probability of a match win of 82% and a match draw of 10% (assuming all 12 games are played)

This should be heavily caveated. In particular

1. I may have misunderstood the probability theory (the probability of misunderstanding probability theory is high)
2. My calculations may be wrong
3. The match finishes when there is a winner rather than going the distance (I'm not sure what difference that makes)
4. Presumably there is an iterative calculation to assess the probability of a Carlsen win on a tie-break - I would guess his chances must be more than 60%
5. Of course, the games are not independent events.
6. No idea if the 60% draw assumption is reasonable.

Overall, that gives about an 88% chance of victory which is about 8/1 on, so the right ballpark, perhaps. Presumably there are people who do this sort of thing for a living, and will do it better.

[NickFaulks]

Presumably there are people who do this sort of thing for a living, and will do it better.

In a relatively unknown field, my guess is that they do it worse.

[IM Jack Rudd]

3. The match finishes when there is a winner rather than going the distance (I'm not sure what difference that makes)

None to the win probabilities. Of course, it changes the probability distribution of final scores.

[JustinHorton]

Karjakin has beaten him just once in 20 games.

I used the same thinking to predict the outcome of a match in 1972

Just out of interest - which world championship contestants have met each other most before their match, and which the least?

[Paul Cooksey]

I think Anand and Gelfand for most, unless we include previous K-K WC matches.

Tal-Botvinnik for least.

[PeterFarr]

Karjakin has beaten him just once in 20 games.

I used the same thinking to predict the outcome of a match in 1972

Just out of interest - which world championship contestants have met each other most before their match, and which the least?

I guess you might want to exclude games between the two players in previous world championship matches, otherwise I guess for the most then the last Karpov v Kasparov match? Otherwise maybe Smyslov and Botvinnik?

Least - maybe Spassky and Fischer? Or Lasker - Capablanca or Steinitz - Lasker?

All just guesswork...

[PeterFarr]

Ah, Paul Cooksey too good for me!

[JustinHorton]

Are there definitely no other nils than Botvinnik-Tal? And what's the actual pre-match Anand-Gelfand tally, does anybody care enough to check?

[Roger de Coverly]

Tal-Botvinnik for least.

Did they even play at all?

Botvinnik always struck me as very much a system player. He would analyse complex themes in the opening and middlegame out to about move 20. He then had to survive twenty moves on his own before, as Bronstein put it, the game was paused for analysis, adjournment in other words.

Perhaps never having played Tal, he didn't know quite what to expect.

[Roger de Coverly]

And what's the actual pre-match Anand-Gelfand tally

I'm getting a total of 69, but that will include some Blitz and Rapid.

Kasparov and Karpov appear to have only met three times before they started the 48 game inconclusive marathon in 1984.

[JustinHorton]

How often had Steinitz played Gunsberg before their 1891 match?

[Roger de Coverly]

How often had Steinitz played Gunsberg before their 1891 match?

Databases can be patchy in their coverage of that era, but that's another contender for the null result.