Shove analysis

Replay Poker Poker Discussion
1

Rather than posting hands themselves, I’m going to start compiling statistics about the hands in a spreadsheet, and once I have a lot of hands, I am going to update this thread with some statistical analysis, and we can see whether the outcomes I’m seeing are at all in line with reality, or if they’re running out so bad that you just have to conclude that curses are real.

To that end, I am asking for advice in this post for how to design the spreadsheet in order to capture the necessary relevant information.

Here’s an example of the data I’m thinking of collecting:

Tournament ID Hand ID Link P1 P2 P3 F1 F2 F3 Turn River Shove street Initiate or Call Covered Reason Hand strength at decision Equity at decision Hand strength at showdown Result Pot Type
656048130 [link] AdQs 7d7s 7h 3c Ac 4h 5s Flop Initiate No Short stack on bubble needs to play aggressively; Strong top pair; Likely ahead vs most hands; Fold equity Top pair AAQ 1.64% Top pair lose -7371 Cooler

Explanation of fields:

Tournament ID: The Replay tournament ID of the hand in question.
Hand ID: The Replay hand ID of the hand in question.
Link: Hyperlink to the hand replay page.
P1, P2, P3: The hole cards held by the players involved in the hand. (There are rarely 4-way shoves, so I think 3 columns is adequate; I can always add another if need be.)
F1|F2|F3|Turn|River: The cards from the flop, turn, and river.
Shove street: Did the shove take place preflop, on the flop, turn, or river?
Initiate or Call: Did I shove or call a shove?
Covered: Did I have enough chips to cover my opponent’s stack?
Reason: Reason(s) I made the decision. Basically a notes field explaining the choice.
Hand strength at decision: self-explanatory
Equity at decision: % to win the pot, if known, based on hand calculator
Hand strength at showdown: self-explanatory; compare to strength at decision to see if I improved or not
Result: Win or lose
Pot: Chips won or lost
Type: The general character of the hand, eg, cooler, rivered, good fold, etc.

This is a lot of information to collect per hand and so will be time consuming to do it for every hand that I am involved with where someone at the table shoves, but I can commit to doing it over a few weeks until I have a nice set of data.

Should I collect more than this? I think this is probably reasonably thorough, but is there other information that would be useful? Obviously the complete hand can be reviewed by going to the link and extracting any additional information desired. I don’t think it’s practical for me to collect the complete hand data including every decision made at every street by every player, but if that were possible to collect in an automated fashion I think there might potentially be value found in that analysis.

I will use this information to count percentage of hands won/lost, profit/loss, how often I shove vs. how often I’m shoved on, How often I get in good vs bad, How often opponents suck out and cooler me, are there any interesting tendencies, such as I tend to win more pots when they’re small vs large, etc.

Using this information, I should be able to run some analysis on my spreadsheet of hand data, and be able to say things like:

“In Tournament 123456, I had my opponent all-in 5 times; I was 50%+ to win the hand in all five, but I only one 1, the final hand, for a 20% win rate; total chips won over these 5 hands was -7500.”

“In hands where I get all-in ahead, I win 52% of the time; probability expects that I should have won these hands 65% of the time. Therefore, something isn’t right about this, but perhaps it’s just that the sample size is too small, and this is still within the margin of error.”

Finally, I’m not sure how to “sum” or “average” or “aggregate” or whatever my combined odds of winning multiple separate hands. Is there a statistically valid way of doing this?

I probably won’t have enough data to make a findings post on this thread for weeks if not months, but I’d like to get the input of seasoned players who have a decent background in math and who are interested in offering suggestions.

2

It might be useful to include your stack size in BBs.

If it was me, I would only look at hands I shoved, not ones I called, and I would only count it if I shoved pre, flop, or turn. I wouldn’t count river jams because the cards are dealt at that point, and no longer involve the pRNG.

I would also only count situations where I only got one caller. This would make data collection easier, if nothing else. I think it would give a better “apples to apples” comparison.

I have no idea how an accountant would aggregate or average the results, but I would look at the results in “bands,” and treat them all as the same equity. You could, for example, say 35% EQ +/- 2%, 40% +/- 2%, and so on, or do 21% to 30%, 31% to 40%, and so on.

The “overs” and “unders” will more or less cancel, and the end results should be pretty close if you do it this way. Not exact, but probably close enough.

3

Thanks for those suggestions, my friend!

I do think that stack size is relevant. It certainly enters into the decision-making. I’ll call wider when I have a giant stack advantage, or when V is down to only a couple of BB. Thanks for that suggestion.

Only tracking all-in hands with two players is simpler, both in data collection and analysis; it’ll make things easier.

For calculating the aggregated probability, I’m still trying to work out how to think about it. I’ll probably post some more questions about this in the near future.

4

So, in order to do this analysis, I need to mathematically define “bad beat”.

Here’s my working definition in progress. What do you think?

A bad beat is a hand that you lose when your are >50% to win at the time you made the decision to shove/call, as determined by a poker hand odds calculator, knowing only the cards at the time of the decision to call is made, and the hole cards are face-up.

Now, I’m not sure about the 50+% threshold; I think losing a coin flip is not necessarily a bad beat. But if I don’t set the line at 50%, I’m not sure where to set it.

I guess we could say that the concept of “bad beat” is more properly thought of as a spectrum, rather than binary – the higher your odds of winning the hand, the more losing the hand is a bad beat. “Bad-beatness” could then be measured on a scale of 50-99+%.

So if you lose a ~50% hand, like a middle pocket pair vs. two overcards, it’s (50/50) = 0% bad beat, and if you lose a 95.5% hand – say you flop a full house, and V sucks out a straight flush on the river – that’s a (45.5/50) = 91% bad beat.

This seems reasonable to me, and easily workable. I think the above is definitely one reasonably good definition for a bad beat (can you suggest any way to improve it?).

But perhaps there are others that can be identified as well.

For example, getting coolered isn’t technically a “bad beat” according to the definition above. But if you constantly get your seemingly strong hands creamed by your opponent hitting miracle coolers, that’s a different sort of bad beat. Coolers will happen, and that’s fine, you should still play your strong hands like they’re strong, and not suspect that you’re going to get coolered every time – but how can I measure whether I’m getting coolered at above the rate expected by probability?

5

Let’s say you will win a situation 60% of the time. This means you will lose it 40% of the time. Why does it matter if you call the losses bad beats, coolered, or whatever? How do arbitrary and ill-defined terms change the math? If you lose there more than 40%, it’s above expectation. If you win more than 60%, it’s above.

For what it’s worth, “bad beats” also involve pot odds, at least to me. For example, I bet the pot and someone takes the 2-1 pot odds to chase down a 5-1 draw. I don’t consider it to be a bad beat if they were chasing with the correct odds. Implied odds muddy these waters some, but still.

Added: The number of beats you take has to be related to your betting for this study to have any real benefit, IMO. That’s the only way you can use your data to take corrective action, otherwise, you’re just looking for evidence of some sort of “curse.”

6

It’s a good question to ask. Right now, I’m not sure it matters.

Poker players have come up with a vocabulary, the terms they came up with have been given meaning. Whether they also have significance, I don’t know. So it’s worth thinking about, and taking a look-see. What will I find out? I don’t know. I’m trying to figure out and understand.

Right now, I have a perception that I lose a disproportionate amount of hands that I’m all-in, or up against an all-in opponent.

If I’m going to test this hypothesis, I need to define my terms.

If I simply call every hand I lose a “bad beat” then I am deluding myself, I can’t learn from it, and of course I won’t become better.

If I analyze these hands and discover that after a second look most of these beats aren’t actually bad, then maybe that helps. I think in order to act correctly, I need to see the world rightly.

So I think it’s worth checking.

7

If my brain is perceiving the world correctly, then in a hand such as:

AQo vs. KQo, if I’m holding AQ, I can expect a King on the board and no Ace about 80% of the time.

According to the hand calculator, I should be about 75% to win. So I should in no way win this hand only 20% of the time. Either my brain is deluding me, or the site is rigged. Analysis should help me to see if the site is rigged. (I doubt it is rigged, but my experiences keep causing me to feel like it is. So I’m trying to debug my perception.)

What’s the actual outcome when I run this hand 10000000s of times on Replay?

If I filter the hands by pot size, is there a noticeable difference in win rate?

If I limp, and V limps, and we check to the river, am I winning more often (Probably; that should mean that neither of us has hit a hand, and I win with A-high).

If I bet huge, is there a King on the board and no Ace far more often than probability tells me to expect? If so, then what’s up with that?

If I raise <10% of my stack size preflop with AQ, does that seem to correlate with winning these hands?

If I raise >25% of my stack, am I losing more? Why? (Because betting bigger folds worse hands, leaving behind only hands that beat you more; that would then suggest that bet sizing is a hole, and I can fix that.) BUT what if I track those big preflop hands, and I’m actually getting beat disproportionately by dominated hands, like KQ, JQ, A4? Then something’s wrong.

8

How will you factor in instances where you shove and everyone folds?

To me, this question points to an obvious problem. You over-realize your equity every time this happens, how will this affect the final results?

For one thing, you will never know your actual equity in these situations. Also, let’s say for example that you jam AA and get JJ to fold. If you should win there 80% of the time, but win 100% of the time when they fold, what does this do to your overall data?

Is collecting a lot of data worth your time if the results are of questionable value?

9

Good question, not sure.

I am thinking about using pot size or net chips won/lost in some way, as well as stack size, but I’m not clear how to use this data.

Well, right. And that’s the purpose of a shove, sometimes.

Short stacked in a tournament, if your stack is still big enough to have fold equity, then you will want to shove preflop in order to steal blinds/antes and limps.

There’s times when this is especially good to do, such as near the bubble, against a table of mostly middle-stacks who are playing tight and trying to get deeper into the tournament/into the money.

It’s best to do it from late position, and in hands where no one has opened ahead of you. If it works, you get 1.5+BB, which at the later blind levels, can be a substantial boost to your stack.

Ideally you do not want to get called that often, and if so, then it may not matter your actual hand strength. It only matters if someone calls. So if you have a good read of the table and know you can get away with it, it’s great.

I of course have successfully advertised on these forums for about two years that I win all-in hands about 5% of the time, so it just doesn’t work. I get called after the 2nd or 3rd shove, or the 1st if I’m so lucky to have detected the premium hand at the table.

When I get a call, I will lose about 95% of the time, and it doesn’t matter what my cards are, or V’s. I can have AQ, they can have 53, and they’ll hit a straight or pair their bottom card while I miss. But usually the dealer awards them an overkill hand where the community cards smash their hole cards for trips or better.

And it’s fine to lose and bust in these situations, I think; it’s better than bleeding out and getting no deeper in the tournament.

This is true, and one reason I suspect I perceive that I get “bad beat” more than I really do.

My hand calculator will tell me my odds against “any two cards”, and that’s all I know until they flip up and I get the bad news. But usually I have a made hand, obviously a vulnerable made hand, and inevitably the caller will have exactly the right cards for the board to be drawing with a nut advantage, or they’ll just flat out be ahead, with set over my two pair, or top pair top kicker, or whatever.

It’s literally bad news every time; it’s bad news that they called, and it portends bad news at showdown.

(I’m trying to avoid injecting too much emotional reaction into how I describe these hands, and be more detached and logical about them, but as you can see, it still creeps in. I can’t help but be sarcastic/cynical about how these hands always run for me.)

I think what I should be doing is calculating my equity not against any two cards, but against a defined range of hands. My phone app that I use for calculating odds doesn’t do that. (I never use the phone app during play, only retrospectively.) Obviously, if your opponent’s range is any two cards, your odds are significantly better than if your opponent’s range is “exactly the two cards that it would make perfect sense to call a shove with here.”

What I don’t understand is how the dealer manages to conspire to see to it that V gets those cards every single time I shove, and then deals them the necessary board cards to be my undoing. It sure feels to me like I’m not getting anywhere near my equity with these hands.

Like, aren’t flush draws supposed to miss sometimes? Like 60% of the time? When does this happen? Just the times I have the flush draw. So the odds look “correct”, perhaps. Flushes fill at the expected rate overall; but the fill rate is skewed against me by a lot, and the only time I actually do fill flush draws is if the pot stays small and V isn’t interested in the pot. When they do VPIP big, then of course that’s when I don’t fill, or when they manage to hit boat or quads. If the site only applies the skew against me in very big pots, then it will be well within the margin of expected variance from probability, and not look unusual. But if I figure out the right filters to apply to the data, maybe I’ll see a glaring, obvious skew happening in these situations.

Is anything I do worth my time? It hardly seems like it. I can just plod on and keep doing the same thing, I guess. Or I can try to observe, measure, understand, and learn.

10

You can calculate a simple “luck factor” with this data. For hands where you got all in vs at least one opponent (either you called or they called, doesn’t matter which):

  1. Count the number of hands you win (count 50% for a 2-way chop etc). Call this W.
  2. Calculate your equity at the point in the hand where you were all in. Add up your total equity across all hands. Call this E.
  3. “Luck factor” is W/E. If the number is >1, you won more than expected, and vice versa.

You can make a nice Excel chart by calculating the 100 hand rolling average of this to see how you’ve been running recently.

For bad beats, you can do a similar sort of calculation:

  1. Count the number of times you get all in as the favorite with >50% equity (or pick some bigger % if you prefer). Call this F.
  2. Count the number of these times that you get beat in these cases. Call this B.
  3. Your bad beat ratio (BB) is B/F.

Now calculate this for times you won when all in as the underdog. This would be your suckout ratio (SO).

Now calculate BB/SO to see whether you’re getting bad beats more than you’re sucking out yourself.

2 Likes
11

Brilliant. Thank you. That’s just what I was trying to think of.

12

There’s an alternative way I’ve been thinking about defining suckout.

1: Calculate equity, and if you’re over some threshold (50+% or whatever you want to decide) and lose the hand, it’s a suck out.

2: Compare the current hand strength at the time of decision, and if you’re ahead there, but behind at the river, that’s a suckout.

For example: a hand like weak Top pair at the flop, up against a hand with a lot of strong draws: two overcards, suited, with straight and flush draws at the flop, but no made hand. Here, you have a ton of outs, but are behind. If V hits one of their outs to win the hand, should that be counted as a suck-out or not? V had more equity due to the number of out cards, but was behind at the time of the decision.

13

After one day:
Hands 31
Hands won 10.5
Hands lost 20.5
Win rate 33.87%
Chips -20569
Suck-out losses 4
Suck-out winners 3
Cumulative equity 1219%
Luck rate (wins/equity) 86%
Suck rate (losses/equity) 168%

See, I’m doing terrible in all-in hands!

33% win rate when either I or my opponent is all-in.

-20.5k in tournament chips net loss.

Looking at the raw data, which is too much to post here, even with 60% confidence (after knowing V’s hole cards) it is insufficient to call a shove – it’s a perennial loser.

It’s only one day, so you might think it’s too small of a data sample to draw any conclusion from, and of course it is – but, this tiny sample is very representative of my experience day after day after day.

14

Yet you have also posted walls of data showing how well you are doing in heads up SnGs. Doesn’t every one of your wins end when you call someone’s allin and win?

15

Might the answer be to exercise more caution when going all in?

16

Kind of.

In the last couple of days I’ve been losing, down about 800k from peak.

Even when I am winning I still run into more than my share of bad beats on all-in hands. I will have to beat a player 4-5 times for each game I win. Each time I have them all in, they double up due to the grace of Small Stack Immunity and the All-in Curse.

I wear them back down and put them at risk multiple times, and eventually win. But very often when I lose a game, I’ll have V down to their last 200-800 chips, and then lose from a 5:1 stack advantage.

If V has more than 800 chips, there’s no way to win against them. The only all-in pots I win reliably are for the final <250 chips. There, I’m closer to .500. Overall, I’m about .334.

17

Sure; I mean, i could auto-fold to all-in and let people walk all over me exploiting that instead. I can fold top pair, two pair, straights on any board with a pair or two or two-suited on the flop. Preflop I can only call with KK+.

To give you an idea of what I go through, last night I played a 100k heads-up game against a player. I had gotten him down to below 2000 chips, then down to 1500. He starts shoving preflop frequently, not every hand, but pretty much every 2-3 hands. I don’t call until the 6th or 7th shove, at which point he’s already regained back to close to starting stack. I call with ATo, he has K3s. So naturally I miss the board while he catches top pair, Kings, doubles up, takes the big stack, and now I’m down where he was, with blinds at 100/200 or 200/400, I forget, and I only get to bleed for a few hands before I’m down to 800 chips, shove, and of course lose another hand where my hole cards dominated his.

This was a game where I had beaten him over 100+ hands, and gotten dealt AA twice, KK, and made quad TTTTens. He would never play a raised pot with me, so every time I had a pocket pair, I only got min-payout with the blinds. Still, over the course of around 100 hands, I had him down to around 1/3 to 1/2 his starting stack, only to have no defense against random shoving.

What’s the point in playing this game if your skill can’t prevail against players if they just resort to brute forcing, shoving a wide range, which they can get away with because the site has pre-ordained that I can never win an all-in hand for more than 3BB?

18

This is him showing caution. Are you saying he outplayed you? That he read you correctly, while you read him wrong? Maybe you aren’t varying your pattern enough to fool him into that final fatal misstep. (Maybe he sold his soul to the devil for good luck, though that seems unlikely.)

When the other guy(s) beats you consistently, there’s a reason. Are you flapping your ears like semaphore flags? Gotta find that leak or tell and plug/fix it. Or, find weaker opponents to take advantage of.

19

This was a weak/passive player who would seldom call a 2BB raise, and who wouldn’t continue past the flop, and seldom bet when he did hit, but would call. I learned quickly that it wasn’t worth bluffing him for more than a 1BB feeler bet, which would fold him unless he had any pair or better. He didn’t win big pots against me, and a few hands where he was willing to call I had better hands and got paid for them, not much but to the tune of 4-6BB.

I don’t fault his strategy, but I had beaten him. Only once he gave up on his weak passive stragegy and turned to random shoving did that change.

I can’t beat random shoving unless I can count on realistic probability. If I won 60% of the hands where I was 60% favorite, I wouldn’t think anything wrong, and would just play with those expectations.

I understand that a 60% hand loses 40% of the time, which is a lot. But I can count on it every big pot, and having a 60% loss rate in 60% favorite situations prevents me from playing winning poker, it makes me feel like the dice are loaded.

I’m hoping that at the end of this analysis, I’ll be able to convince myself that it’s not so.

20
Stats
Hands 87
Hands won 37
Hands lost 48
Win rate 42.53%
Chips -63302
Suck-out losses 11
Suck-out winners 13
Cumulative equity 3810% 43.79%
Luck rate 97%
Suck rate 126%

Observations:

My win rate actually closely agrees with my average equity.

I actually suck out a few more times than they suck out on me.

Yet, I manage to lose -63302 chips.

I’m under 100% in luck (winning when flipping up shows I’m behind), and 126% in suck (losing when I was ahead at the time the hand went face-up).

So, what to make of this? Losing so many chips despite realizing my equity “about right” sure feels like some seriously bad luck.

We can still say the sample size of the number of hands is tiny, of course. But this is absolutely representative of how I run for weeks and months on end.

I can keep tracking numbers, and I’m sure unless I figure out how to play differently, these trends will continue in their present direction.

However, i can short circuit all that work, and propose the following, if anyone wants to help me out in an experiment.

I’m looking for someone to meet me at the smallest-stakes 2-seat NLHE ring table available. We both buy-in at the standard (minimum) buy-in. Looks like that’s a 10-20 table at 1000 chips. We go all-in-preflop every single hand for 20 hands. Loser re-buys at the standard buy-in, and we keep going like this. After 20 hands, if I haven’t won at least 30% of the pots, I will delete my account and quit playing poker on this site forever.

That’s 6 hands out of 20, and you’ll never have to hear me complain ever again.

Come on, I want to get out of here. Just all-in and call 20 times

Who’ll take that action? Any player who plays regularly at 500/1000 it’s like playing for the blinds, with no further betting, winner goes to whoever shows down the better hand. A small price to pay to win 14000 chips AND get rid of me forever.