Luck adjusted winnings in SNGs

[Voyeurism]

Great. Will I have to reimport those hands/tournaments or will they update automatically?

[WhiteRider]

I'm pretty sure you'll need to reimport, although if it is only prize detection you may be able to correct it by running Auto Detect on your tournaments via the Edit Tournament Results window.

[Voyeurism]

I've updated to the latest (4.01.6) beta, created a new database, and reimported my hand histories for January. This is the resulting "Results by Tournament" tab:

I think there are still problems with the ICM/Equity values being calculated, and this screenshot shows that there are possibly also issues with the currency conversion.

If you look at tournament #508073664 (the 4th line down, highlighted in blue), in which I finished second receiving $8.39 total, you can see that "My C Net Won" is £3.11 and "Net Won" is $4.89. Prize money $8.39 - Buyin $3.50 = Net Won $4.89, so this figure is correct. The figure of £3.11 implies an conversion rate of £0.636/$ (which, without checking, seems like a reasonable conversion rate).

Moving along the row, "My C Net Adjusted" is £4.27 which is higher than "My C Net Won", while "Net Adjusted" is $2.99 which is lower than "Net Won". The implied conversion rate here is £1.428/$ (which, without checking, seems like the wrong conversion rate).

Clearly something is wrong here, and it seems like its simply to do with the currency conversions. However, as I finished second in this STT, my "Net Adjusted" for the tournament (which I understand is calculated solely based on the last hand) cannot possibly have been less than my "Net Won" for the tournament - even if the money went in with me drawing dead, I cannot possibly end up with less than second place money. The minimum equity should therefore have been $4.89.

This is particularly concerning because, as this hand was heads up, $EV = CEV and no "ICM" should be required - this implies that the computation of CEV may also be off, which (I assume) could be causing errors in other parts of the program.

The hand history in question is in the spoiler, and as an aside my own calculation of my equity in this hand is $11.47 when the money goes in - spreadsheet calculation attached (zipped, as .xlsx is not allowed).

Spoiler: show

Code: Select all

PokerStars Hand #74657391295: Tournament #508073664, $3.11+$0.39 USD Hold'em No Limit - Level VI (100/200) - 2012/01/29 3:20:13 WET [2012/01/28 22:20:13 ET]
Table '508073664 1' 9-max Seat #3 is the button
Seat 2: DPenman (4080 in chips)
Seat 3: MrKovi (9420 in chips)
MrKovi: posts small blind 100
DPenman: posts big blind 200
*** HOLE CARDS ***
Dealt to DPenman [Js Jd]
MrKovi: raises 500 to 700
DPenman: raises 3380 to 4080 and is all-in
MrKovi: calls 3380
*** FLOP *** [3c 5s 9d]
*** TURN *** [3c 5s 9d] [Ks]
*** RIVER *** [3c 5s 9d Ks] [As]
*** SHOW DOWN ***
DPenman: shows [Js Jd] (a pair of Jacks)
MrKovi: shows [Ad Kc] (two pair, Aces and Kings)
MrKovi collected 8160 from pot
DPenman finished the tournament in 2nd place and received $8.39.
MrKovi wins the tournament and receives $14.01 - congratulations!
*** SUMMARY ***
Total pot 8160 | Rake 0
Board [3c 5s 9d Ks As]
Seat 2: DPenman (big blind) showed [Js Jd] and lost with a pair of Jacks
Seat 3: MrKovi (button) (small blind) showed [Ad Kc] and won (8160) with two pair, Aces and Kings


[WhiteRider]

Could you please attach the files you imported to a Support Ticket (a selection including these examples is fine) and post your ticket number here - I'll make sure the development team see this thread.
Thanks.

[_dave_]

Just updated to the new version PT4, imported January tournaments (all turbo HUSNG on Stars) and the EV adjusted winning line is quite wrong. Probably it's doing "2" that Tarix described above, but I do not understand why that would ever be implemented, it's completely irrelevant? can anyone explain why on earth the "final hand" is of relevance, if say you got stacked and left with a chip and a chair the hand prior? Seems just as valid to take the equity of the seventy-fourth hand, even if you folded pre - as in, lol wat. Equity adjusted tourney winnings are the sum of every all-in. Maybe I'm missing something, but why would you do this please?

Also add results graph in buyins as a matter of utmost importance. consider it comparable to "winnings". it's far more important than "chips" and "bb", which are already present in the "winnings in" drop-down.

It's also still calculating equity very "loosely", but I's sure that's on the to-do list already.

Some great fixes in this update though, most obviously to me in the last half hour to the tourney detection - not a single error

[Yogi Rob]

Just updated to the new version PT4, imported January tournaments (all turbo HUSNG on Stars) and the EV adjusted winning line is quite wrong. Probably it's doing "2" that Tarix described above, but I do not understand why that would ever be implemented, it's completely irrelevant? can anyone explain why on earth the "final hand" is of relevance, if say you got stacked and left with a chip and a chair the hand prior? Seems just as valid to take the equity of the seventy-fourth hand, even if you folded pre - as in, lol wat. Equity adjusted tourney winnings are the sum of every all-in. Maybe I'm missing something, but why would you do this please?

For you, as a HU player, it's preferable to have stats as similar as possible to cash. There is no perfect way of calculating luck adjusted winnings for tourneys. For you, ICM means very little, for a DON tourney player, ICM means everything, and for most tourney players ICM dictates strategy.

I've considered using a rolling stack based on every hand EV and using the final theoretical stack ICM as the Adjusted, but that opens up some other problems. Since the theoretical stack won't resemble the real stack, there will be lots of cases where the "real stack" strategy would directly conflict with the "theoretical stack" strategy. Also, your theoretical stack would affect the other players' stacks. For the vast majority of tourney players where ICM optimal, tight-early aggressive-late strategy rules the day, that final hand is incredibly important, and many STTs are 1-hand games. We are also neutralizing the luck factor of final hands that you're not even involved in- for an extreme example, in a DON you're on the bubble as a "chip-and-a-chair" and two giant stacks get it all in with AA vs AA (incorrectly), and player X wins with a flush. Player X got lucky against player Y and you won because of that luck. Final hand ICM based adjusted neutralizes the luck factor in this example, where the situation might not have existed at all in the theoretical world where all stack sizes would have been different.

However, your input is very important to us, and I agree that for HU, the rules are very different. I will discuss this with the other developers. We may be able to do something.

Also add results graph in buyins as a matter of utmost importance. consider it comparable to "winnings". it's far more important than "chips" and "bb", which are already present in the "winnings in" drop-down.

I assume you mean x-axis would represent buyin money rather than tourneys and y-axis winnings? We will take this into consideration.

It's also still calculating equity very "loosely", but I's sure that's on the to-do list already.

Some great fixes in this update though, most obviously to me in the last half hour to the tourney detection - not a single error

Thanks

[Yogi Rob]

Clearly something is wrong here, and it seems like its simply to do with the currency conversions. However, as I finished second in this STT, my "Net Adjusted" for the tournament (which I understand is calculated solely based on the last hand) cannot possibly have been less than my "Net Won" for the tournament - even if the money went in with me drawing dead, I cannot possibly end up with less than second place money. The minimum equity should therefore have been $4.89.

Yes, you are correct. This will be fixed top priority. This looks like some currency conversion is being done when it shouldn't.

This is particularly concerning because, as this hand was heads up, $EV = CEV and no "ICM" should be required - this implies that the computation of CEV may also be off, which (I assume) could be causing errors in other parts of the program.

Not sure if I follow here. ICM is still used in "Net Adjusted" because prize distribution is 50%/30% in this HU situation. ICM is not used in chip EV.

Thanks for finding this. [edit: fixed for next release]

[Voyeurism]

I've considered using a rolling stack based on every hand EV and using the final theoretical stack ICM as the Adjusted, but that opens up some other problems. Since the theoretical stack won't resemble the real stack, there will be lots of cases where the "real stack" strategy would directly conflict with the "theoretical stack" strategy. Also, your theoretical stack would affect the other players' stacks. For the vast majority of tourney players where ICM optimal, tight-early aggressive-late strategy rules the day, that final hand is incredibly important, and many STTs are 1-hand games. We are also neutralizing the luck factor of final hands that you're not even involved in- for an extreme example, in a DON you're on the bubble as a "chip-and-a-chair" and two giant stacks get it all in with AA vs AA (incorrectly), and player X wins with a flush. Player X got lucky against player Y and you won because of that luck. Final hand ICM based adjusted neutralizes the luck factor in this example, where the situation might not have existed at all in the theoretical world where all stack sizes would have been different.

I would have thought the easiest method for calculating ICM adjusted all-in equities for a given STT is:

1. Assign a "starting EV" for the tournament - because the STT is raked this would be a negative figure, and assuming all players are of equal skill would be directly proportional to the tournament fee as a percentage of total prize pool.
2. For each hand in which an all-in pot occurs, calculate:


(a) the actual (ICM) value of hero's stack at the start of the hand
(b) the expected (ICM) value of hero's stack at the end of the hand (distributing chips in the pot according to the participant's chip equity when the money went in)
(c) the difference between the two - that is, (b)-(a)
(d) write (c) to the database - (a) and (b) are not strictly necessary
3. If no all-in pot occurs for the hand, the value of (c) is 0 - that is, no change.
4. Sum the values of (c) for each hand in the tournament, and add to the "starting EV" to get the expected tournament equity.

This is particularly concerning because, as this hand was heads up, $EV = CEV and no "ICM" should be required - this implies that the computation of CEV may also be off, which (I assume) could be causing errors in other parts of the program.

Not sure if I follow here. ICM is still used in "Net Adjusted" because prize distribution is 50%/30% in this HU situation. ICM is not used in chip EV.

All I meant by this is that once you are heads up, the value of each chip becomes constant rather than varying based on the number of other chips in your stack. Although the prize distribution is 50%/30%, in a very real sense it's actually 100%/0% of a different prize pool because each remaining player is guarenteed to leave with second place money, and the amount for which players are competing is the difference between first and second place money (in this case, 20% of the original prize pool). Thus having 50% of the chips in play can be assumed to give you a 50% chance of winning the tournament, and 50% equity in the new prize pool (in our example, this would be equal to 10% of the original prize pool).

In this situation, $EV is directly proportional to CEV and you can do all the calculations without needing ICM - it's basically just a heads up cash game where chips are worth ((first place prize + second place prize)-2(second place prize))/(total chips in play) each, and a single calculation in pokerstove is all that's necessary if you're doing it manually. That's probably not the best way to explain what I meant, but I hope you get the gist of it.

Thanks for finding this. [edit: fixed for next release]

I assume this means you no longer require me to create a support ticket - very efficient.

[Voyeurism]

Another thought regarding the "final hand only" adjusted Net: it seems like this method of calculating adjusted winnings makes it very hard for a player to run "above EV".

I suspect that the following reasoning is likely way off, but surely the only "final hand" situation in which your actual winnings exceed adjusted winnings (ie. you run "above EV") is when you are the winner of the tournament? Every other tournament outcome involves you losing the "final hand" (because it's your final hand, not the final hand of the tournament), which necessarily results in your actual winnings being equal to (money went in drawing dead) or less than (money went in drawing live) your adjusted winnings.

Like I said, I suspect that this is crazy talk but I can't see why.

[_dave_]

I would have thought the easiest method for calculating ICM adjusted all-in equities for a given STT is:

1. Assign a "starting EV" for the tournament - because the STT is raked this would be a negative figure, and assuming all players are of equal skill would be directly proportional to the tournament fee as a percentage of total prize pool.
2. For each hand in which an all-in pot occurs, calculate:


(a) the actual (ICM) value of hero's stack at the start of the hand
(b) the expected (ICM) value of hero's stack at the end of the hand (distributing chips in the pot according to the participant's chip equity when the money went in)
(c) the difference between the two - that is, (b)-(a)
(d) write (c) to the database - (a) and (b) are not strictly necessary
3. If no all-in pot occurs for the hand, the value of (c) is 0 - that is, no change.
4. Sum the values of (c) for each hand in the tournament, and add to the "starting EV" to get the expected tournament equity.

Pretty sure this is the HEM1 method, and is the correct way of doing tournament EV.

There is no perfect way of calculating luck adjusted winnings for tourneys. For you, ICM means very little, for a DON tourney player, ICM means everything, and for most tourney players ICM dictates strategy.

I played DoNs for a good while before PS removed them - I will make a database and import them for science sometime soon - but no - "final hand ICM" is imo going to be very misleading for DoNs too. It seems it wil always come down to a coinflip of whether you cover your opponent or not. you cover him, you can make a huge ICM blunder (calling multiple way allin with 32o for 95% of stack), for no result - then make a great call with AKs for you last BB, and final hand EV will say you got "unlucky". Well, I guess the EV line will stay kinda flat since you had so few chips, wher it should plummet.

There does exist a perfect method for luck adjusted winnings in all single table tournaments, and that is the HEM1 method probably described above. for each all in hand, calc EQ for the hand matchup and multiply by chips won/lost to get cEV-Difference. At the end of the total, sum cEV-Diff - final stack gives you what you "should have won" in chips. fanagle that with the prizepool and you get a perfect result (easy for HU lol).

In hindsight I imagine "final hand Equity" could be useful for showing something for MTTs SNGMTTS, I don't play them really so can't be sure. I'd imagine it's still gonna be flawed, but for some people maybe it's better than nothing? for STTs though, sum ICM adjusted Equity for every all-in hand is the "gold standard". There's a massive thread in STTF on 2+2 going back years where every angle has been attempted to be poked at by the "I don't believe the EV line" crowd, and aside from where hand history formats have caused errors it's well accepted there are no issues with the calculations.

I assume you mean x-axis would represent buyin money rather than tourneys and y-axis winnings? We will take this into consideration.

No, the Y-axis, where it shows currently Prizes, chips or BBs - this needs to show "Buyins" also - graph is same as normal one, just no $$ showing.