I am the World's Greatest Unknown WSOP Historian... ask me anything (especially before 2003)

The answer is (or at least was prior to 2017) 9-10

With, interestingly enough, offsuit winning more hands than suited

Aces is often what people presumed, and it definitely won many hands also, despite often losing to hands like 910

We studied this (and then I taught it) at each University I was at

The lesson wasn't so much about being results oriented with saying that you should always play 9-10 because it won more hands than any other 2 card combos

The lesson was about how EVERYONE played premiums like AA- and psychos loved stuff like 27+ but hands specifically like 9-10 would win a bunch of times due to how infrequently they'd be played and how often they'd either be folded on the flop or wind up winning at showdown or to a fold

During the Gus Hansen High Stakes Poker years, all the suited connectors and then suited one gappers dramatically increased in their frequency of times they were held post-flop, which lowered their effectiveness from the pre-Moneymaker years when folks like the DNegs/Juanda/Cunningham/Ivey Crew took advantage of playing them

It is a useless Trivia fact that got lost to the Ether when WSOP changed their storage/server priorities but I'm basically Cliff Clavin so I've never forgot how odd it was from a statistical standpoint along with other notable standouts like 7-10 and, as you mentioned, 6-8 👍

Fascinating, I would have guess Jack-Ten in that case. I'd still like to know where your data comes from (I hope you're not including just televised hands). Also, it sounds like you're talking about how often the hands won at showdown as opposed to what percentage of times they won as hole cards period.

And why offsuit over suited (unless, again, you're only talking about showdown, in which case I can see opponents folding to the obvious flushes)?

J10 ranked decently high but seems to have caught too many losses vs 10Q+ as well as lower cards (like A79 board calling flop vs 77 or 99)

The company contracted to track hand histories from 1998 to 2015 provided the (anonymized but hard to obscure late stage televised hands) data directly to us for Game Theory course review

They presumed there wouldn't be much value in the data other than to illustrate basic statistics and they were right

Obviously as per the agreement, we'd have to provide them with any actionable/useful information and would've been subject to their NDA on all of it if that happened, since they owned the usage and the tournament owned the IP, etc. (but that didn't happen)

There was some basic Solver stuff in there but it wasn't even as abundant or useful as a small team of multitablers grinding online sites to get a larger sample size so that didn't really matter either

Fwiw, the veracity of the input data began to lag around 2010 and so I had no interest after 2015

I wouldn't even know what the current state is as we were never contracted directly back then anyway lol

To your other question, as originally stated, "winningest" just meant the most hands won by a starting hand when a flop occurred

To your point, there was always a bit of blind spot on exactly what was involved and whether any hands where a flop happened but there was no showdown were included

Or how they handled early stages where no hole card cameras existed and/or times when players would obscure from cameras or readers, whether they won or lost without showdown

I, personally, can only attest to how 2005-2015 was processed and was told to simply accept what the original team was given in 1999 for all previous years lol

Obviously this weakly compiled "data" wouldn't stand up to basic peer review so we'd made it clear that anything worthy of publishing (in a peer reviewed journal) would have to come from a more rigorous undertaking in a future year's ME

Even as Solvers (not us) started earning enough money to buy ad space, our contacts made it clear they had no interest in letting us make the data collection process more scientific and rigorous

So it became nothing more than a short portion of a single day's lecture on statistics in traditional community games (which you probably know is not really what Game Theory is about anyway) each year and the equivalency of maybe 2 or 3 pointless bar stool trivia "facts" (not so much about 9-10 specifically but just that Aces were never #1 due to factors like human nature and being a game of chance)

I guess that's what I expected. The sample of hands was very biased towards the more interesting hands (or at least what ESPN considered interesting; I don't agree with their overweighting of all in hands).

That's a huge consideration. Most Ten-Nine hands were folded before the flop but weren't counted as losers.

It all makes sense now.

Oh, not really sure as there wasn't an opportunity to get into a lot of potentially interesting queries like that

Just that some years the specificity of suit was included or would read J10o vs J10s for offsuit vs suited and it just so happened that offsuit won slightly more than half of the times that any 910 combo won

Could very well be because of what you said in terms of folding to something like a suited ace or even a single ace (or king) on a 4 or 5 flush board

More likely given the nature of what went into the numerator and denominator would be that most Main Event players would seek more rundowns with suited 910 vs offsuit and end up not winning those hands

(So 910s, like a lot of other suited hands, would increase the number of flops seen in losing hands.)

I don't remember some other useful examples of outcomes like that off the top of my head other than 27o seeing more flops than 27s for whatever reason and many hands 10Q- seeming to be almost random in terms of whether they won more often suited vs offsuit. Suited Broadways saw more flops than unsuited as did certain clusters like 35, 67, 68.

We'd asked 2 of our Psych professors if they had thoughts on that but they said it'd be nearly impossible to determine such things outside of a controlled lab experiment with extensive sample sizes and diverse participant pools.
Which would mean costly.
Which would mean they assumed no one would ever fund something like that. Lol

Exactly.
And 27o saw many more flops than hands like 28s or 39s because fans of televised poker with Seven Deuce bounties are adrenaline junkies or whatever. Lol

There was also no weighting in terms of either SPR or level of Blinds or even shorthanded late stages vs full tables.

So hands like 48s or 2Ko that get immediately folded on Day 1 suddenly see flops later on.

Not exactly a robust set of "data" to work with.

But the sheer volume of the number of times that any 910 would win was absurd.
Someone (allegedly) told Lon M (in passing) about how much more successful 910 had been than Jacks and he'd said something like "I'm not surprised" 😆