Analysis - Playstyle Changes Based On Lead

When you're ahead in a game of Mahjong, do you call more to try to finish hands faster and move the game along? Do you stay dama instead of calling riichi to increase your safety? Do you deal into hands that look cheap? Let's see what the Houous do.

In the Average Houou Player blog post, we looked at the average rates for the players. Call rate is 34.58%, riichi rate is 16.93%, deal-in rate is 12.86%, and win rate is 20.87%. The average win is 5638. Let's look at how this changes for the first place player based on how wide their lead is.

The "Lead" column is in buckets. The <8000 row doesn't include the data from the <4000 row, so it's more of a 4000~7900 row. But, writing it this way is more concise. The number is the difference between first place's score and second place's score, ignoring the other players.

They call a bit more while in first with a small lead. The riichi rate declines sharply as it goes down, which of course brings the average win value down. While I was running the analysis, I noticed that the riichi rates in the early replays were lower than the ones in the higher replays. So, let's look at what these numbers are for the earliest 600,000 replays I have.

Comparing it with the overall one above, you can see the difference in riichi rates. Now let's look at the latter 600,000 replays for comparison.

In recent times, people are more willing to riichi. The meta has shifted. Looking between the riichi and dealin columns, the increased riichi rate doesn't have that big of an increase in deal-in rate associated with it and the win rate has gone up.

You can see that the deal-in rate goes up in the >20000 row. This could be from players doing sashikomi, a technique where they try to intentionally deal into cheap hands to advance the round. You can see this from the average deal-in value going down at the same time. Let's compare the deal-in distribution for the <4000 case and the >20000 case.

There are a lot more 1000 or 2000 deal-ins with an overwhelming lead.

You might also play differently based on whether you're in first as the dealer or as a non-dealer. A dealer with a huge lead won't get as much benefit from calling, as it doesn't move the game along. Let's see how the numbers change based on whether the player is the dealer.

The call rate is actually higher with a small lead. In flat situations, it's easier to justify calling as a dealer, since your hands are naturally worth more. Similarly, dealing in isn't as bad, since you won't be dealing into a dealer, and the difference between dealing in and an opponent tsumoing isn't as big, since you pay extra on tsumo anyway. It's interesting to see how high the dealer riichi rates are.

The data can be found here.