Mima Ito vs Hina Hayata Finals Preview: A Statistical Approach
Not the post you were looking for? A guide to all of Edges and Nets’ coverage of WTT Doha (also known as WTT Middle East Hub and formerly known as ITTF Qatar Open) can be found here.
Mima Ito and Hina Hayata will be facing off in the finals of WTT Contender at WTT Doha shortly. To get ready for the finals match-up, Edges and Nets re-watched their seven game thriller at the All Japan National Championships this January.
We took a new approach of first qualitatively looking for trends and then performing a brief quantitative analysis to confirm our intuition based on manually labeled data. The game went a full seven games for a total of 129 points, giving us a decent sample size. Overall, Ito won 48% of the points despite winning 4-3; this is because she lost a couple games by a wide margin and won all her own games narrowly.
We present two conclusions below. The first conclusion we verify is pretty obvious from watching the tape, but the second insight may be non-obvious without actually looking at the numbers. Although the analysis is primitive, we hope that this post provides a glimpse of a future with more automatically labeled data, from which we may be able to quickly draw further insights from particular match-ups without having to watch too much film.
Longer Rallies Favor Mima Ito
One trend that jumps out when you watch the National Championships matches is that Hina Hayata commonly wins points by exploiting Ito’s height and putting Ito out of position within the first shot or two, and then finishing the point one shot later. However, once Ito is able to get in position in the rally, she is able to return pretty much anything Hayata throws at her.
To verify this idea, we manually labeled the shot-length and the winner of each rally. Shot length was measured by the number of shots Mima Ito attempted to make (e.g. if Ito misses her third-ball attack or wins the point on her third-ball attack, it is considered a 2-shot rally either way). The reason we count shot attempts and not made is that Ito makes an extra shot when she wins a point, thus biasing the results if we count made shots.
Due to limited sample size, we do not want to perform too fine grained analysis, we divide points into “short” points in which Ito attempted 3 or fewer shots (i.e. the total rally was at most five or six shots depending on who served) and “long” points in which Ito attempted 4 or more shots (i.e. the rallly was at least six or seven shots depending on who served).
87% of the points were considered short points, which makes sense since a lot of table tennis is executing your service and service returns well. In other words, on average there were two and a half long rallies each game. This can absolutely swing the match, as Ito won two games 11-9 and two games 11-8.
Ito only won 44% of the short points, but won a staggering 71% of long points, including all six rallies in which Ito attempted five or more shots. Obviously there is some noise due to small sample size and potential unknown source of bias in our approach, but the results are quite stark.
We thus highlight the importance it is for Hayata to be able to finish the point quickly, although that is obviously easier said than done.
How Should Hayata Manage Her Long Serves?
One of the key challenges in playing Mima Ito is managing long serves. Probably the worst serve one can make when playing Mima Ito is a short serve to her backhand, as that gives her free reign to do whatever combination of banana and strawberry flicks and short and deep pushes that she likes with her short pips. As a result, opponents typically avoid essentially completely avoiding this serve.
The two good serves to Ito are the short serve to the forehand, which prevents her from getting creative with the short pips without getting slightly out of position, and the long serve to the backhand, which forces her to give a predictable and softer return. However, the long serve carries risk, since when Ito anticipates it coming, she can step around for a hard forehand opening against the long serve. Serve too many times long to the backhand, and one may end up simply asking to be killed by her forehand.
It is not obvious just from watching the film which serve is more effective, and it likely varies by match-up and the opponent’s ability to execute each serve. However, we can draw some insight for Hayata by performing quantitative analysis on Hayata’s last match with Ito.
Ito won 52% of the points on her own serve and 44% of the points on serve return, which sounds about reasonable. Hayata served long on roughly one third of her serves, and we can assume that the remaining two-thirds were short and to Ito’s forehand.
The sample size is small as our splits are quite fine-grained, but the results are somewhat interesting. Hayata won 55% of the points in which she served short and to the forehand. On the other hand, she won 60% of the points in which she served long. At least in the previous match, it appears that serving long yielded better results for Hayata than serving short.
It may seem that Hayata should be serving long more often, but we have to consider that the more Hayata uses them then the more Ito will start stepping around, which would decrease the efficacy of the long serve to the backhand but increase the efficacy of the short serve to the forehand.
When looking at the splits between when Ito received the long serve with her forehand or her backhand, this tradeoff appears to emerge: Hayata won seven out of thirteen (54%) of the points that Ito took with her forehand but six out of the nine (67%) points that Ito returned with her backhand.
However, the sample size is tiny, so we cannot draw any strong statistical conclusions; all it would take is one edge ball from Ito to make the efficacy of a serve to the backhand to only be 55%. If the results do hold on larger data, it matches our intuition (and apparently Ito’s since she keeps stepping around) that Ito returns the long serve better with her forehand. If that is the case, left-handed opponents like Hayata may want to consider serving more often from the center of the table in order to land the wide serve to Ito’s backhand.
On the flip side, if after analyzing more data it appears that taking the serve with the forehand and backhand yield similar results, Ito may want to consider if she wants to step around less often, which would presumably allow her to focus more on the short forehand return. That would be a surprising and counter-intuitive result for many including Edges and Nets, but we have seen large-scale quantitative analysis upend common intuition in various other sports.
The data used in the analysis of this post was both primitive and small in scale, but we hope some of the conclusions that we drew offer a glimpse of what can happen in the future given enough well-labeled data. The length of rally, length of serve, and whether a player used forehand or backhand should actually all be pretty easily trackable based on modern AI techniques, and it will be exciting to see what the future holds for quantitative analysis in table tennis.