Takao Shinji's 900 Wins and Assessing Pro Performance

[ez4u]

When Takao beat Iyama in December to take the Tengen title it was his 899th professional win. The following week on Christmas Day, he held on to win a half-pointer against Yamashita in the Meijin League for his 900th win. He is the youngest Nihon Kiin professional to hit this landmark at 38 years and 1 month (18 months younger than Cho Chikun did it), has the best winning record percentage in doing so at 70.0% (versus 68.2% for Kobayashi Satoru), and managed it in the shortest professional career of 23 years and 8 months (versus Yoda Norio's 27 years and 2 months). Congratulations to Takao!
However, we know from an earlier post by JF that Yuki Satoshi has had an even more impressive career in this regard, leading up to his 1000th win a couple of years ago.
But what does it all mean anyway? I particularly did not know what to make of their winning percentages. The actual speed to accumulate wins will be affected by the number of games available, which have grown in recent years. So what will affect the winning rate? One factor will be their respective strength. These days we have a way to compare their strength using mamumamu0413's ratings of Japanese professionals. With a bit of drudgery I was able to assemble a comparison of their ratings at the same age throughout their careers. The original graph ran up to age 42 (Yuki's current age). The comparison of Takao and Yuki yields an ambiguous answer since at different periods they each have sustained a higher rating than the other. So I added Cho Chikun's ratings through age 42 to the graph. This makes things more interesting because he clearly maintained a higher rating in virtually every year of his career while achieving a lower winning ratio.

Yuki Takao ChoC ratings by age.jpg (45.23 KiB) Viewed 5869 times

So what else will affect their winning ratios? The other element is the quality of their competition. This idea led in two directions. The first is in regard to Yuki, who is a member of the Kansai Kiin. Is there a difference in ratings between the Nihon Kiin and the Kansai Kiin pros? The second graph below is a breakdown of the current (January 2015) ratings table comparing the 317 Nihon Kiin pros to the 115 Kansai Kiin pros by dan and in total. The light and dark colored bars are joined at the average rating and extend one standard deviation up (light colors) or down (dark colors). The average Nihon Kiin pro has a higher rating in every dan rank except 2p. Overall the Nihon Kiin pros have a rating of 6.15 versus 5.55 for the Kansai Kiin (see the 'All' columns on the right). This can go a long way to explaining Yuki's very high career winning rate.
(More on the second factor related to opponents' strength later. I have to run out right now!)

Nihon Kiin vs Kansai Kiin ratings by dan.jpg (47.69 KiB) Viewed 5869 times

[ez4u]

The second item that I wanted to introduce here is the different level of competition found at the different stages of the Japanese tournaments. There is a nice table on mamumamu0413's site reporting the results of the last ten years in the seven major tournaments in Japan (Kisei, Meijin, Honinbo, Oza, Tengen, Gosei, and Judan). The table shows a very clear progression in the rated strength of the participants as we climb the ladder from the earliest qualifying tournament to the title match itself. This is entirely expected and reassuring since it shows that the process of choosing strong challengers is working. But the opposite side is that the more successful a player is (and Cho Chikun is clearly the most successful of the three players we are looking at, in terms of number of titles won) the tougher opponents they face on average and the lower winning rate we would expect to see in support of their rating. This would help explain why Cho had a lower winning rate in achieving his 900 or 1000 wins compared to Takao and Yuki.

Code: Select all

Level of Play                        Ave. Rating
Title match winners                         9.04
Title match losers                          8.93
League members                              8.62
Challenger selection tournament members     8.24
Final qualifier members                     7.87
A qualifier members                         7.22
B qualifier members                         6.44
C qualifier members                         5.72

Since we have this table in front of us, it is worth mentioning that the A, B, and C qualifiers shown here are separated geographically between Tokyo and the combined Nagoya (Chubu) and Osaka (Kansai) areas. As such only the pros who make it into the final qualifying tournaments and beyond play against combined pools of players from both the Nihon Kiin and Kansai Kiin. The 229 Nihon Kiin members in Tokyo will play the early qualifiers against each other, without meeting either the Kansai Kiin pros or the 88 members of the Nagoya and Osaka branches of the Nihon Kiin. The rating data shows that the Tokyo pros will face stronger opponents on average than those playing in Nagoya and Osaka. This should lead to lower winning ratios for Tokyo pros compared to equally rated players in the other branches and the Kansai Kiin.

[oren]

Can you add Iyama as a comparison for the first few years against Tako, Yuki, and Cho? It would be interesting.

[ez4u]

That is really the subject of a separate post on age-related development. I have pulled a bunch of numbers off the ratings site and am currently trying to put it together in a visually sensible way. RSN!

[snorri]

I almost had a heart attack when I saw the line for Cho on your graph and thought you were talking Cho U and had a horrible vision that Cho U was 42. Of course, you meant Cho Chikun and I'd have spared myself the pain if I'd read the text before the graph.

I like your other graph on the performance vs. professional dan rank. I knew that pro dan ranks were getting to the point of being largely irrelevant, but that graph pretty much sums it up. What's going on with the Kansai Kiin 6p lot?

[ez4u]

... What's going on with the Kansai Kiin 6p lot?

There are 15 6p's at the KK with an average age of 58, which is much older than the comparable group at the NK with an average age of 44.