TGP for unfinished strikes formats?

TGP of strikes tournaments has been determined by simulation for a while now. An issue we’ve been having is that pretty much all the strikes tournaments in our region are not being played until the end, because there’s a separate finals portion, meaning the strikes knockout is played only until all but 4 or 8 or whatever players have been eliminated.

In those cases, TGP is determined after the fact by actually played rounds, but that’s not ideal, it would be nice to be able to communicate the TGP when announcing a tournament, and then actually be able to make good on that promise, and not have the TGP randomly change by how the strikes play out.

Here’s my suggestion: Since TGP is determined by Keefer’s simulator for the default formats, we could just use that to run a simulation and send its results with the tournament registration. If it says it takes 10 rounds on average to whittle down the field to less than 8 players, we would get those 10 meaningful games, no matter if it turns out later that it actually took 9 or 11 rounds.

This is not about exploiting anything, we would get less meaningful games than actually played as often as we would get more. It’s only about being able to construct a tournament format with a predictable and guaranteed TGP.

@pinwizj? @MET?

I don’t believe the simulation included how many players were left at every point in time. Even if it did, it would become extremely cumbersome to list every possible player combination along with the various “If you stop at 8 players = X”, “If you stop at 6 players = Y”, “If you stop at 4 players = Z”.

If you’re consistently doing the same format, your players should have a pretty good idea as to what the value of the event will be. A difference of 10% in either direction because things ran slightly shorter or longer than average isn’t something I would consider material enough to try and estimate for every version of a strikes format for every possible player cutoff point.

I must say that whomever came up with the idea… it’s a good one to use to push a large H2H strikes tourney over the top of the 100% mark. The one I participated in Fulda during my visit there was a 5-strike H2H, and using the most recent version of that from Pinball Olympics, it looks like the 111-player tourney would have had 21 Total Games Played, at max – if you played out the stopping point when it got down to 8 players remaining. Instead, running the PAPA-style finals after Round 16 means 12 (6x2) TGP are added on, getting to 28.

Granted, I’m guessing the PAPA-style finals on 2 moderns and a classic probably took 1.5 hours per round, adding 3 hours onto the tourney after round 16. Compared to the estimated additional 5 rounds of H2H which were taking 25 minutes per round, or around 2 hours. Though the TGP/hour efficiency of the PAPA finals is around 25% better, getting 12 TGP in 3 hours, vs the last 5 rounds of H2H getting 5 TGP in approx 2 hours.

Looking at another one from Olympics 2019, I’m not sure why @Tobstar79 graded it at 84%, as I think it should have been higher at 96%: 12 rounds of actual H2H, plus the 12 TGP from an 8-player PAPA-style finals. = 24 TGP. Or 96%

Comparing to what the TGP would have been had they simply played out the 4-strike H2H to its natural conclusion, the TGP would have been only 15 TGP – a MASSIVE difference!

Thus, my original point and applause to whomever came up with this.

what I really like about strikes followed by a finals component compared to other h2h qualifying formats like round robin/swiss is that every game played is really relevant and you don’t play your last game and it is about finishing 20th or 40th place. And not playing the strikes tournament until the end is plus for me bc you don’t end up in situations where f.e. there are 3 players left and one lucky player gets a more or less random bye to the final 2.
The only “downside” is if you’re location is running on coin drops.
Afaik I’m approving results since 2020, so I can’t say anything about the downgrade, but most probably it should be 96%

I’m not so much suggesting that the IFPA add every possible combination to the TGP guide, more that tournament organizers get the option to provide their own numbers in the tournament details, using the official simulator.

E.g. for an 80 players 4-strike head-to-head knockout, with PAPA-style finals for the final 8, an organizer could add the following to the tournament details:

FairStrikeSimV2.exe -p80 -s4 -r10000 -g2 -c 0 1 0 0 1 0 0 0 0
round 13: 6.456100

Meaning on average it takes 12 rounds to get to 8 or less players, so 12 meaningful games would get locked in for a total of 24. It shouldn’t be too much of a hassle for @MET to run the posted command himself when approving the results in case he wants to double check the numbers.

Just to be clear about my intentions here: You know how people can get about their WPPRs—if you’re aiming for 100%, but in the end it’s only 92%, there’s going to be some players blaming the organizers for somehow having botched it. Especially if there would have been enough time to play more, but the strikes just played out in such a way that it randomly just didn’t happen. Communication often isn’t as easy as it is in the US when different countries and languages are involved.

I want to shield my tournament organizers from that sort of bs by locking in the expected value from the get-go, just like organizers who choose to play their knockouts vanilla style get to do.

I guess I don’t see the point of a strikes tournament with a finals especially if you are trying to get a guaranteed number of rounds. Why not just run Matchplay? Is it the advantage of trimming the field for time?

Strikes is typically the last tournament of a multi-tournament weekend, people can start their travel home as soon as they drop out, there’s no real harm in some people dropping out early if they have to catch a flight, etc.

As to why run it with a finals, our players prefer it that way, there’s an attainable achievement you can work towards besides just getting a high rank, a well defined finals portion makes for a better stream, etc.

1 Like

I’m not sure if @MET knows how to run the simulation. I don’t believe @PressStart does and I know I don’t. I also don’t believe it’s as easy as saying it takes 12 rounds so it’s automatically 24 meaningful games played. Looking at the spreadsheet of results that I used to create the TGP guide and @keefer provided an avg. weight that took into account the number of 4-player, 3-player and 2-player groups that each simulation calculated. For example in the chart below, averaging 13 rounds of play could be anywhere between 19 and 24 meaningful games played.

Now if there’s a way to create a user interface out of that program that allows us on the review side to type in # of players, # of strikes, style of strike, # of players remaining, and it spits out the “Meaningful Games for TGP” calculation, then I’m all for it. I could replace the entire TGP guide page with just that calculator and it would make life a ton easier.

Until then, for those areas of the world where English isn’t the first language, perhaps we can have a area at the top of the TGP guide that states in every different language - “Do not worry so much about TGP. Have fun! If the expected TGP and actual TGP differ by 10% I promise that everyone will be okay.”

6 Likes

That’s not what I meant there, I was talking about a head-to-head knockout, so 12 expected rounds would be an expected 12 meaningful games, plus another 12 for 2 rounds of PAPA-style finals, for a total of 24. Not sure if anyone runs group knockouts with a separate finals portion, but Keefer’s command line tool should do a fine job of providing the appropriate weight there.

If that’s on the table, I’ll see if I can whip up a web interface that uses Keefer’s tool in the back end.

1 Like

That’s exactly the format that one of our local TD’s is running. Smallish turnouts so they’re just running group strikes(last place only receives a strike) until they are down to 4 players. Then they do a three game finals PAPA style. I’ve been helping him with calculating TGP since it seems the TGP they’ve been given has been all over the map.

The only sure way to do it is to look at every round and determine if the single game played should get 2X, 1.5X, or just 1X.

This will be fixed after the next rebuild :slight_smile:

Alright, here it is: https://strikestgp.slapsave.com/

It’s just a dumb UI that uses Keefer’s tool on the server side. I had to make minimal changes to the code, though, in order to enable that not-playing-until-the-end feature—I’m happy to have @keefer check the diff to make sure I didn’t mess anything up.

I didn’t put a lot of format presets in there, but any additional format can be set manually in the UI. Obviously, I can easily add more presets if it’s useful. @pinwizj please let me know if you could see something like this replace the precalculated tables. I suppose something will have to be done for 2023 anyway, because all those tables are capped at 25 meaningful games right now.

One thing I noticed while coding this is that in the current TGP guide, all head-to-head knockouts are probably undervalued: The TGP guide assumes that each time there’s an odd number of players in an H2H, there’s going to be one 3-player-group, with 2 of those 3 players getting a strike. In reality, people are using matchplay.events for their knockouts, which assigns a bye in this case, so only 1 of those 3 players is actually getting a strike.

I wrote the UI to match the TGP guide for now, but it might make sense to switch it to the way H2H knockouts are actually being played in the wild. In some cases that difference is as much as 3 meaningful games. This makes adding a separate finals to an H2H knockout a lot more valuable in the current system, because then the tournament gets judged by the actual number of games played instead of the lowball guide value.

1 Like
  1. That’s super fucking cool! I’ll have Adam, Michael and Brian check this out and see what they think, but I love the idea!

  2. Brian is actually working on a way to automatically process results that are run through MatchPlay so it won’t even hit the review process. It will be auto-approved based on the data that he’s able to pull directly out of MatchPlay. What happens to the tournament in MatchPlay when you ‘stop’ at the final 8? Does that tournament just stay open forever?

1 Like

I think you can close it even though it’s not done. Just freezes everyone where they currently stand.

Thanks. I added more presets, so now every strikes format that’s listed in the TGP guide can be selected.

Not sure how exactly the MatchPlay API works, but it might not even matter whether the tournament has been closed or not. If it does, I’m sure @haugstrup will be able to tweak the API so you can get the data you need. The bigger issue here is probably that in all the strikes tournaments I’ve played in, that had PAPA-style finals, those finals have always been done manually and not in MatchPlay, so full automation won’t be possible there anyway.

Generally speaking, I can totally understand the wish to automate this sort of thing, it does unfortunately run counter to what I actually wanted to achieve here, though. Right now, no-finals knockouts have their TGP guaranteed in advance, while with-finals knockouts don’t. I wanted to get that guarantee for the latter, too, but it seems that with full automation looking only at the rounds actually played, neither would have it in the future. I mean, you could of course just use simulator results instead of rounds played even with full automation, but I’m guessing that’s probably not the plan.

I added an option for head-to-head knockouts to choose between how many games the TGP guide will give you, and how many rounds you can actually expect to play when you’re using matchplay.events.

Interestingly, the larger the finals portion, the less difference it makes. If you play a finals with the top 8 remaining players, it’s almost the same in many cases. If you play knockout until the end, it can be a difference of 3 games.

A possible exploit in the current system would be to never play out the last 2 remaining players, and instead do a separate 1-game finals.

The plan is actually to read the format data of the MatchPlay submission and through that process spit out the estimated TGP value based on those variables. We’re not reading actual MatchPlay results for this (if I understand what Brian is planning correctly).

2 Likes

Ah cool, so that would address the issue. You’re probably still going to need some manual work for knockouts with a separate finals, because MatchPlay won’t tell you whether it was supposed to be played until the end or only until n players were left (and if you go by how many players were actually left in the results, you’ll overvalue it).

Anyway, if it turns out you have any use for the the simulator UI, just let me know, I’m happy to adapt it or whatever to make it useful. Same goes for Brian, in case he thinks that for what he’s planning it might make sense to interface with the JSON API I wrote for the simulator, which enables TGP calculation for unfinished knockouts and fixes the issue with head-to-heads being calculated differently than how MatchPlay actually plays them.

1 Like

I’ll always provide Brian with whatever data he wants, but really what I’m wondering here is why you’re not playing your PAPA-style finals through Match Play :joy:

2 Likes