Peak Go Performance Age

[pookpooi]

In chess, there's study among grandmasters pointing out that 31.4 years old seems to be peak performance age. (http://rdcu.be/y2sV)

In Go, according to http://www.goratings.org the youngest peak performance age for famous professional player is Lee Changho (20 years 1 month), and the oldest peak performance age is Sugiuchi Masao (62 years 8 months). Though Sugiuchi play much lower number of games than Rin Kaiho, which peak at 48 years 2 months.

Other big name pro peaks are as follow
Otake Hideo 37y 10m
Kato Masao 39y 4m
Kobayashi Koichi 37y 6m
Nie Weiping 36y
Ma Xiaochun 30y 11m
Yoo Changhyuk 30y 5m
Cho Chikun 25y 9m
Cho Hunhyun 36y 6m
Zhou Heyang 32y 8m

*I only look into pros that are older than 40 years old.

Is the single golden age number exist for Go? The safe bet would be the same number as chess, but without a serious study on this subject we'll never know for sure.

[luigi]

Fluid intelligence (brain computing power) decreases with age, while crystallized intelligence (accrued knowledge) increases with age. I think peak performance age is higher in Go simply because it rewards accrued knowledge more than chess and punishes calculation oversights less than chess.

[gowan]

How is peak performance defined? I can't believe that Cho CHikun's peak was age 25. Surely peak performance extends over a span of years. Fujisawa Shuko must have been at peak when he won the first Meijin tournament at age 37 and continued winning titles up to winning the first the Kisei title at age 51 and holding it for 6 consecutive terms.

I also have doubts about the chess study. There were grandmasters such as Emmanuel Lasker who held the world championship for the longest time in history.

[Kirby]

I heard an interview with Dr. Adam Gazzaley, who is a neuroscience researcher studying ways to enhance cognition.

He described a talk he gave several years ago where they explained the results of a study they did, which basically said, mental performance declines with age, and that the trend is pretty much universal. He recalled that the audience seemed somewhat depressed at the unfortunate result.

At that point in his career, he decided that humans don’t have to accept that result; instead of simply observing how the brain declines with age, why not identify ways to improve it? He began looking into ways that games and brain exercises can improve cognitive ability with (despite?) age.

The idea is that, as humans, we don’t have to accept the current condition.

—-

That being said, some research suggests that attempts so far to enhance cognitive ability through games, etc., are ineffective. One such example is cited here.

So perhaps a decline in cognitive ability is inevitable... Or maybe it’s not.

I hold onto the hope that cognitive decline can be prevented. How that happens, though, I don’t know.

Engaging in mentally challenging activities on a regular basis seems better than doing nothing.

Maybe I’ll get better at go when I’m older. I’ll keep holding onto that hope for now.

[pookpooi]

How is peak performance defined?

My definition is the peak of the player's graph in Goratings.org like this https://www.goratings.org/en/players/98.html

Peak point and ranking is quite a different matter, just as ranking and title winning is a different matter too.

[Calvin Clark]

One challenge with these statistics is that it's a moving baseline with respect to the strength of the field as a whole. Professional go players are getting stronger. They have better problems, can play more games because of the internet, can share information more quickly and (very recently) are able to play against strong AI.

Myungwan Kim 9p said in one lecture during the AlphaGo / Lee matches that he actually believed that Lee Changho 9p was stronger today than before. That's an astonishing and bold statement, but maybe not impossible. It's just that his competition is even better than when he was younger. Of course, they had his games to study, whereas he had to play them first.

[John Fairbairn]

There are several problems with using the goratings criteria. Lack of games is one.

For Cho Chikun, given in the graph above, goratings scores him on the basis of 1799 games (which probably also includes unofficial games) for a score of 60%. But earlier this year Cho reached 1500 wins in official games with a score of 65%. That seems like wide discrepancy to me.

Goratings has under 80% of Cho's games, but almost certainly well under that percentage for other players. Cho is exceptionally well represented because of the efforts of Jan van Rongen in collecting his games.

You can also consider the quality of the games he was playing rather than the number of rabbits he beat. For example, taking the period 1996-1999 (rather than the 1982 of goratings) Cho's title haul was 20th Kisei (1996), 21st Meijin (1996), 51st Honinbo (1996), 43rd NHK (1996), 29th Hayago (1996), 2nd JT (1996), 21st Kisei (1997), 22nd Meijin (1997), 52nd Honinbo (1997), 22nd Kisei (1998), 23rd Meijin (1998), 53rd Honinbo (1998), 23rd Kisei (1999), 24th Meijin (1999). He was thus playing then very many games against the very best opposition. His win rate possibly took a dent for that reason, but his title haul was impressive.

Of course it was also impressive around 1982, but taking a similar 4-year span around then his title haul was 5th Meijin (1980), 6th Meijin (1981), 36th Honinbo (1981), 7th Meijin (1982), 37th Honinbo (1982), 20th Judan (1982), 4th Kakusei (1982), 1st Shusai (1982), 7th Kisei (1983), 8th Meijin (1983), 30th NHK (1983). Am I alone in thinking the later haul glitters more? More of a peak?

[Uberdude]

Lee Changho (20 years 1 month)
Sugiuchi Masao (62 years 8 months)
Rin Kaiho peak at 48 years 2 months.

Other big name pro peaks are as follow
Otake Hideo 37y 10m
Kato Masao 39y 4m
Kobayashi Koichi 37y 6m
Nie Weiping 36y
Ma Xiaochun 30y 11m
Yoo Changhyuk 30y 5m
Cho Chikun 25y 9m
Cho Hunhyun 36y 6m
Zhou Heyang 32y 8m

Pookpooi only looked at the older pros, I thought I'd look at top players of the last decade as I think most of us feel pros are peaking/burning-out younger. Of course maybe they've not yet reached their peak and there's the various caveats and problems with this methodology, but here's a few for comparison:

Code: Select all

Cho U            28 in 2008 dramatic rise and fall
Gu Li            25 in 2008 (2nd peak in 2012) definitely peaked, similar double peak to Lee (plus some steps on his rise up)
Kang Dongyun     19 in 2008 not much rating change to 26 in 2015, recent fall
Kong Jie         27 in 2009 fell quickly 
Lee Sedol        27 in 2010 (2nd peak in 2015 at 32) does seem like his best years are now over after loss to Ke Jie in 2016 MLily
Choi Cheolhan    26 in 2011 (not fallen much, but doubt he will peak again)
Fan Tingyu       16 in 2012 dropped 30pts but definitely a peak from when he won Ing
Shi Yue          22 in 2013 (does seem like a peak, he was Chinese #1 and won his only world title back then), still a top player
Chen Yaoye       23 in 2013 ditto 
Zhou Ruiyang     21 in 2013 ditto, fairly flat to 2016, recent small drop
Kim Jiseok       24 in 2014 almost back to peak now
Park Yeonghun    30 in 2015 his first peak was aged 20 in 2005-2007 when he won first international titles, but had a 2nd bigger peak recently (no wins, but finals of LG and Chunlan)
Ke Jie           20 in 2017 (maybe not a peak)
Park Junghwan    24 in 2017 not peaked
Iyama Yuta       28 in 2017 not peaked
Shin Jinseo      17 in 2017 not peaked

It does seem like the Chinese now peak in their early 20s whereas Koreans (and Chinese 10 years ago) in late 20s. The peaks do pretty well match up when they won international titles for the recent one-hit wonders.

[Bill Spight]

One challenge with these statistics is that it's a moving baseline with respect to the strength of the field as a whole. Professional go players are getting stronger. They have better problems, can play more games because of the internet, can share information more quickly and (very recently) are able to play against strong AI.

And there are more of them. The competition among pros has increased, as well as the competition to become a pro. 100 years ago Shusaku gave 3 stones to low level pros. The field has narrowed. If there are several top level pros competing with each other, their individual ratings peaks will largely be noise.

[pookpooi]

100 years ago Shusaku gave 3 stones to low level pros. The field has narrowed.

This make me dig back to this thread viewtopic.php?f=13&t=12691 to argue with you about the gap among pro.
But I found interesting comment instead that is not off-topic, in the last comment

I just wanted to add this chart from mamumamu's site. It shows the average change in rating by age in his domestic database. On average, pros in Japan have stopped adding rating points at about the age of 30, held relatively steady until about 40, and declined more rapidly from then on. Both ends of the chart have only a few people in the numbers but from 20-50 at least he is averaging across more than 100 (and sometimes nearly 300) people.

Look like the peak point very similar to chess? (At 31.4 years) At least among Japanese pro though

[luigi]

It shows the average change in rating by age in his domestic database.

Average change per game? Surely it's not per year.

[hyperpape]

It shows the average change in rating by age in his domestic database.

Average change per game? Surely it's not per year.

Per year. It’s roughly a ten point scale, so .1 is already a noticeable change.

[hyperpape]

The peak goratings score depends on the range of professionals included, which leads to high ratings in later years. I believe players are initially guessed to have a rating of 3000, so increasing the number of weak players increases the range, pushing the top players to higher numbers.

It is impossible to do historical comparisons using that data, though you might get passable results using points from the top or ordinal ranks for the top players.

[pookpooi]

I believe players are initially guessed to have a rating of 3000

In a very rare occasion, Remi Coulom finally come out and defend his ranking methodology.

"WHR does not really give a start rating to players. That's the big advantage of the WHR approach over incremental algorithms such Taeil Bai's: with WHR, ratings of the past are corrected retroactively based on current games. For incremental rating algorithms, the choice of a good initial rating is very important, but not for WHR.
Players with very few games do get a rating, but it is very uncertain, and does not influence opponents much. And it will be corrected fast if it turns out to be very wrong."

For the full conversation, with most of the questions attacking on how his rank put Japanese pro in high order, can be read here https://www.reddit.com/r/baduk/comments ... _rankings/

For me it feels like the real reason the gap gets bigger is because more games are played, not more players throw into the pool. Yes, it's kinda counter-intuitive that professional gap getting bigger with time when almost every Go article point out that today's gap between professional is narrower than the past. I'm wondering how much this will affect my 'real objective' here, I hope it's still doable. But with all these complication I'll never ever do longitudinal study in my life again.

[bugcat]

By way of comparison, I decided to apply John's method of examining title wins in order to identify peak periods to Kobayashi Koichi.

From what I can tell, his peak period was a lengthy five years from 1988 to 1993, which corresponds to about 35 - 40 years old (which nowadays we'd consider past one's prime). In every one of these years, he won the Kisei, Gosei, Meijin, and (apart from 1992) the China-Japan Meijin.

The peak period of Takemiya Masaki looks to be 1985-88, the period in which he had his main reign as Honinbo. That corresponds roughly to the ages 34 - 37, similar to Kobayashi.