Originally Posted by sauce123
Tl;DR
Cliffs
-I think make-up staking is exploitative and terrible
-I present a system I think is a lot better with a lot of boring math and discussion of incentives
I've staked 50+ people over 8+ years and I hope my learning experiences can be of use to the community.
Staking at make-up is terrible
Make-up staking is a system that sometimes works, but that badly mis-aligns the interests of horse and staker when deep in make-up. The misalignment of incentives creates a lot of negativity and pain. When deep in make-up, 100% of profits (at least on paper) go to the staker. What actually happens is that a staker will put a horse in a big game (or the horse hits a bad run, or plays very bad, etc) and the horse gets in so much make-up that they can't see daylight. The horse feels like they're a slave, and won't get any profit for months or years. If the staker is greedy, they'll apply psychological pressure and guilt in order to enforce the contract and receive 100% of their profits until makeup is cleared. If the staker is kinder, they'll reach a profit sharing agreement so that the horse can live a somewhat healthy lifestyle while they claw back to even. In some cases, the staker will forgive all or some of the makeup, and even pay the horse some profits while in makeup, so that the horse can continue to be motivated and play their best. In practice, there's always some kind of complex negotiation going on where strong emotions like guilt, fear and despair predominate for the horse, and greed, anger and guilt predominate for the staker, and both parties grow to resent each other.
Even though the staking with MU system is standard and often works, I've spent years avoiding it like the plague because it feels so bad. After much tinkering and reality testing, here's what I've come up with.
Alternative: Tiered action sharing with Mark-Up
Of all the systems I've experimented with, the only one that works reliably is what I'll call 'Tiered action sharing with mark-up', for brevity TASMU (which sounds like a brand of hair gel or something but **** it).
It works like this: say I'm backing a $2/$5 cash game player to play $5/$10 through $10/$20, and to take shots at an occasional soft $25/$50 game. As a baseline, I require that they take $5 big blind worth of their action at every stake- so they have 100% of themselves at $2/$5 (their current game), 50% at $5/$10, etc. Secondly, I increase their share of their own action at each increasing stake so perhaps they have 60% of themselves at $5/$10 ($6 big blind), 7/20th of themselves at $10/$20 ($7 big blind), and $8.50/$50 of themselves at $25/$50 ($8.50 big blind). The point of this is that if they're playing bigger on my money, then they're playing bigger on their own money. The rate of their own increasing action in TASMU is variable and takes skill on the part of the staker to set correctly.
Next, I set a decreasing mark-up as stakes increase on the portion of the action that I have. Elaborating on the previous example, if I have 40% of their action at $5/$10, and they have 60%, then I'll set some mark-up on the action such that they're incentivized to play $5/$10 and give me action even if a decent $2/$5 game is available. Perhaps this number is 2bb/100. So, the horse's total incentive structure at $5/$10 is 60% of their own action, and selling me 40% of their action at 2bb/100; in the case of a live game that gets 30 hands/hour, then that'd be $2.40/hour cash payment. Setting the size of the mark-up on the side action to align your interests with your horse's takes skill and attention to detail, and a knowledge of game conditions.
At $10/$20, (again elaborating on the previous example), the action is split 7/20th horse and 13/20th staker. To properly incentivize the horse to play $10/$20 over a good $5/$10 or $2/$5 game, I'll pay mark up on my 13/20th share. I have to pay less mark-up at $10/$20 than $5/$10 because the volume of marked up action is $13bb instead of $4 bb. Perhaps instead of 2bb/100, I pay 1.4 bb/100. Setting this second mark up in relation to the 0th and 1st takes skill and a knowledge of game conditions.
Finally at $25/$50 I might set a mark up of 1bb/100 on my $41.50 of the action. This results in an hourly of .3*$41.50= $12.45 for the horse in addition to the horse's $8.50/$50 of the action at face. Relating this markup to the prior ones takes skill and knowledge of game conditions.
Periodically, I renegotiate in good faith with my horse as their bankroll grows. At some point they become a $5/$10 player and I start getting action at $10/$20. That's to be expected, and that's the whole point.
What can go wrong?
The key variable to consider here as a staker is keeping the horse's EV generated from mark-up well balanced with their EV from their own action. In other words, if they choose to play bad and try to cash in on the mark-up money, then it's crucial that they lose even more of their own money so that they can't game the system. If they can't game the system, incentives are aligned.
This is most important at the highest stakes available in the deal, in our example, $25/$50 where the horse has $8.50 bb themselves and is paid 1bb/100 in markup. For simplicity, I'll ignore the hands/hour conversion and just use bb/100 for both winrates and mark ups.
There's a small window where the system can be gamed, and that's when the EV from mark up is greater than the horse's lossrate on their own action. In the example, horse is being paid 1bb/100 in mark up on 83% of the action, or $41.50/100. If the horse's piece of their own action plus their mark up payment is EV >=$0, the horse can gain while the staker loses. So, since horse has 17% of their own action, this ratio is 83/17, or ~4.88. In the limit then, horse's EV is 17%*-4.88bb/100*$50+$41.50MU=0, and staker's EV is (83%*-4.88bb/100*$50)-$41.50=-$244/100. The loss limit for the staker in this structure is then ~5bb/100 at $25/$50 in the example. The maximum value for the horse while the deal is still 0EV for the staker is also important to know because this value will represent the marginal value to the horse to play in a softer game which is likely more +EV for staker. This value is of course +1bb/100, and EVs at that winrate are [staker] (.83*1bb/100)-(.83*1bb/100)=0, and for [horse] (.17*1*50)+(.83*1*50)=+$50.
What protects the staker in these cases is that it is not a good hourly for the horse to play in low/negative winrate games that don't provide the horse a better hourly than an available alternative, and also that games falling in the misalignment window don't come around very often, and aren't usually that costly even when they happen and horse decides to play in them. In games where EVs are low and the horse is winning $50/hour and staker is winning $0/hour, the hope is that there are better games in the room such that the horse would never want to play in such a low winrate game. It can also be helpful to have a candid conversation between staker and horse, and to build the level of trust such that a horse won't play big in marginal games that are reasonably likely to fall in the misalignment window.
Be very careful setting the size of this window where horse can profit at staker's expense!
So, I think a skillful setting of mark-ups and exposures in TASMU can almost entirely align incentives for horse and staker, and avoid much of the pain of make-up staking. At least in the real world, where there's a scarcity of games and personal relationships can smooth over the perverse incentives falling in the misalignment window.
I've so far ignored the massive topic of non financial incentives that come into play. It's entirely possible a horse simply enjoys playing bigger games, and will play big for fun, losing both their money and their staker's money. It's also entirely possible the horse or staker (or both) are bad at rationally predicting winrates in varying game conditions, and will come to an agreement that's exploitative due to a lack of skill at predicting winrates, exposures and markups. There are tons of ways this can go wrong besides the cut and dry world of financial incentives which I can't possibly list here. Staking is very tricky and dangerous for both parties, trust and communication are always important, and I caution anyone from engaging in these practices without very careful consideration.
I also haven't discussed the term, or termination conditions, of a TASMU deal. What's important in this context is that they aren't “until you quit poker,” as is the case in make-up deals. I've found 6months to 2 years is a useful guideline, depending on the situation. Early termination conditions for the horse might include some small lump sum payment, or might occur only on certain conditions stipulated in an agreement.
I've also ignored how to stake a player who is flat broke. TASMU won't help with this.
At bottom though, I've found TASMU is a much better way of structuring workable staking deals that provide value to both parties, with a lower potential for exploitation. I have been both a horse and a staker in TASMU deals, and have felt OK on both ends. I hope others can make use of this system, and that it helps some avoid repetitions of the sort of situation described in this thread, or at least lessens the pain of extricating oneself from a staking deal gone bad.