Subsidizing Prediction Markets

Link post

Epistemic Sta­tus: Sadly Not Yet Subsidized

Robin Han­son linked to my pre­vi­ous post on pre­dic­tion mar­kets with the fol­low­ing note:

I did briefly men­tion sub­si­diza­tion, as an op­tion for satis­fy­ing the fifth re­quire­ment: Sources of Disagree­ment and In­ter­est, also known as Suck­ers At The Table. The ul­ti­mate sucker is an ex­plicit, in­ten­tional one. It can serve that roll quite well, and a suffi­ciently large sub­sidy can make up for a lot. Any suffi­ciently large sucker can do that – give peo­ple enough profit to chase, and sud­denly not be­ing so well-defined or so quick or prob­a­ble to re­solve, or even be­ing not safe from key in­sider in­for­ma­tion, starts to sound like it is worth the risk.

Sup­pose one wants to cre­ate a sub­si­dized pre­dic­tion mar­ket. Your goal pre­sum­ably is to get a good es­ti­mate for the prob­a­bil­ity dis­tri­bu­tion of an event, and to do so with­out pay­ing more than nec­es­sary. Se­condary goals might in­clude build­ing up in­ter­est and a mar­ket­place for this and fu­ture pre­dic­tion mar­kets, and get­ting a trans­par­ently ro­bust re­sult, so oth­ers or even the me­dia are more likely to take the out­come se­ri­ously. What is the best way to go about do­ing this?

Be­fore look­ing at im­ple­men­ta­tion de­tails, I’ll look at the five things a pre­dic­tion mar­ket needs.

I

Well Defined

The most cost-effi­cient sub­sidy for a mar­ket is to en­sure that the mar­ket is well defined. Some­one has to make sure ev­ery­one un­der­stands ex­actly what hap­pens un­der ev­ery sce­nario, and that some­one is you. Care­ful word­ing and con­sid­er­a­tion of cor­ner cases is vi­tal. Tak­ing the time to do this right is a lot more effi­cient than throw­ing money at the prob­lem, es­pe­cially try­ing to build a sys­tem and brand over time.

If you’re go­ing to sub­si­dize a mar­ket, step one is to write good care­ful rules, make sure peo­ple un­der­stand them, and to com­mit to mak­ing it right for ev­ery­one if some­thing goes wrong, if nec­es­sary by pay­ing mul­ti­ple sides as if they had won. This is po­ten­tially quite ex­pen­sive even if it rarely hap­pens, so it’s hard to bud­get for it, and it feels bad in the mo­ment so of­ten peo­ple don’t pull the trig­ger. Plus, if you do it some­times, peo­ple will ar­gue for it all the time.

But if you’re in it to win it, this is where you start.

II

Quick Resolution

Once you’ve got your defi­ni­tions set­tled, your next job is to pay the win­ners quickly once the event hap­pens. Peo­ple care about this more than you can pos­si­bly imag­ine. The differ­ence be­tween pay­ing out five sec­onds af­ter the fi­nal play, and five min­utes af­ter the fi­nal play, is a big game to many. Make them wait an hour and they’ll be fu­ri­ously com­plain­ing on fo­rums. When the out­come is cer­tain, even if it hasn’t ac­tu­ally hap­pened yet, it’s of­ten a great move to pay peo­ple in ad­vance. Peo­ple love it. Of course, oc­ca­sion­ally some­one does some­thing like pay out on bets on Hillary Clin­ton two weeks early, in which case you end up pay­ing both sides. But great pub­lic­ity, and good sub­sidy!

Another key ser­vice is to make sure your sys­tem rec­og­nizes when a profit has been locked in, or risk has been hedged, and does not need­lessly tie up cap­i­tal.

This one is oth­er­wise tough to work around. If you want to know what hap­pens twenty years from now, noth­ing is go­ing to make re­solv­ing the ques­tion hap­pen quickly. You can help a lot by en­sur­ing that the mar­ket is liquid. If I buy in at 50% now, then a year from now the mar­ket is at 75% and is liquid enough, I can take my prof­its in one year rather than twenty. That’s still a year, and it’s still un­likely the price will ‘catch up’ fully to my new opinion by that time. It helps a lot, though.

III

Prob­a­ble Resolution

It is a large feel bad, and a real ex­pense, when cap­i­tal is tied up and odds look good but then the event doesn’t hap­pen, and funds are re­turned. It hurts most when you’ve pul­led off an ar­bi­trage, and you win money on any re­sult.

If, when this hap­pens, you sub­si­dize peo­ple for their time and cap­i­tal, they’d be much more ex­cited to par­ti­ci­pate. I think this would have a stronger effect than a similar sub­sidy to the mar­ket it­self, once you get enough liquidity to jump start trad­ing. Make sure that if money gets tied up for months or years, that it won’t be for noth­ing.

IV

Limited Hid­den Information

If your goal is to buy the hid­den in­for­ma­tion, you might be all right with oth­ers not be­ing in­ter­ested in your mar­ket, as long as the sub­sidy brings the in­sid­ers to the table to mop up the free money. That ap­proach is quite ex­pen­sive. If the reg­u­lar traders are still driven away, you’ll end up pay­ing a lot to get the in­sid­ers to show their hand, be­cause they can’t make money off any­one else. Even in­sid­ers start to worry that oth­ers have more and bet­ter in­side in­for­ma­tion than they do, which could put them at a dis­ad­van­tage. So it’s still im­por­tant to bring in the out­siders.

One ap­proach is to make the in­side in­for­ma­tion pub­lic. Do your own in­ves­ti­ga­tions, re­quire dis­clo­sure from those par­ti­ci­pat­ing in the events them­selves, work to keep ev­ery­one in­formed. That helps but when what you want to get at is the in­side in­for­ma­tion it only goes so far.

That means that when this is your prob­lem, and you can’t fix it di­rectly through ac­tion and dis­clo­sure, you’re go­ing to have to spend a lot of money. The key is to give that money to the out­siders as much as pos­si­ble. They are the ones you need at the table, to get your­self a good mar­ket. The in­sid­ers can then prey on the out­siders, but that’s much bet­ter than prey­ing on you di­rectly.

The coun­ter­ar­gu­ment, es­pe­cially if you don’t need to show liquidity or vol­ume, is that if you buy the in­for­ma­tion di­rectly there’s less noise, so per­haps you want to de­sign the sys­tem to get a small num­ber of highly in­formed traders and let ev­ery­one else get driven away. In cases where the out­siders would be pure noise, where the in­sid­ers out­right know the an­swer, and where get­ting out­siders to be suck­ers that take a loss isn’t prac­ti­cal, that can be best.

V

Disagree­ment and Interest

This one’s easy. You are pay­ing a sub­sidy, so you’re the sucker. Be loud about it so ev­ery­one knows you’re the sucker, and then they can fight to cash in. Ex­cel­lent.

The other half, dis­agree­ment, is still im­por­tant. Many peo­ple, whose anal­y­sis and par­ti­ci­pa­tion you want, still benefit from a story that ex­plains why they are be­ing paid to ex­press an opinion, rather than fight­ing to be slightly more effi­cient at cap­tur­ing the sub­sidy. And of course, if no one dis­agrees about the an­swer, then your sub­sidy was wasted, since you already knew the an­swer!

In light of those is­sues, what are the best ways to sub­si­dize the mar­ket?

Op­tion 0: Cover Your Basics

Solve the is­sues noted above. Choose a mar­ket peo­ple want to par­ti­ci­pate in to be­gin with. En­sure there are care­fully writ­ten rules with no am­bi­guity, that any prob­lems there are cov­ered. Make sure you’ll get things re­solved and paid quickly, that cap­i­tal won’t be tied up one minute longer than nec­es­sary. When pos­si­ble, dis­close all the rele­vant in­for­ma­tion, on all lev­els. If things don’t re­solve, it would be great if you could com­pen­sate peo­ple for their time and cap­i­tal.

And also, make sure ev­ery­one is con­fi­dent the win­ners will be paid! Noth­ing kills a mar­ket like wor­ry­ing you can’t col­lect if you win. That’s of­ten as or more im­por­tant even than pro­vid­ing strong, re­li­able liquidity.

If you can im­prove your in­ter­face, us­abil­ity, ac­cessibil­ity, user’s tax li­a­bil­ity or any­thing like that, definitely do that. If your mar­ket de­sign is poor, such as hav­ing the wrong tick size, make sure to fix that. Tick sizes that are too small dis­cour­age the pro­vid­ing of liquidity to the mar­ket, and are in my ex­pe­rience a big­ger and more com­mon mis­take than ticks that are too big.

Fi­nally, waive the fees. All of them. De­posit fees, with­draw fees, trad­ing fees, you name it. At most, there should be a fee when tak­ing liquidity that is paid en­tirely to the trader pro­vid­ing liquidity. Peo­ple hate pay­ing fees a lot more than they like get­ting sub­sidies. They won’t can­cel out.

With that out of the way, what are your op­tions for the main sub­sidy?

Op­tion 1: Be a Mar­ket Maker and Provide Liquidity Directly

As the sub­si­dizer of the mar­ket, com­mit to be­ing the mar­ket maker with well-defined rules.

The stan­dard prin­ci­ple is, let ev­ery­one know that there will always be $X of liquidity available on both sides, and at a fixed cost of Y% price differ­ence be­tween your bid and your offer. So for ex­am­ple, you might agree to offer $1,000 on each side with a differ­ence of 5% at all times, start­ing with a 48% bid and a 53% offer. You’d then ad­just as you did trades.

A sim­ple rule to pro­tect your­self from un­limited down­side is if you do a trade for some per­cent of your liquidity, you ad­just your price that per­centage of its width. So in this ex­am­ple, if some­one took 40% of your offer, you’d ad­just by 40% of 5%, which is 2%, and now have a 50% bid and a 55% offer. If you fol­low such a rule, your max­i­mum loss is what it takes to move the odds to 0% or 100% (and if you let peo­ple keep trad­ing un­til the event is done, you will take that loss). Peo­ple trad­ing against you in op­po­site di­rec­tions can make you money, but can’t cost you money.

For con­ve­nience, you can post ad­di­tional bids and offers so that if some­one wants to move the odds a lot, they can see what liquidity they would get from you, and have the op­tion to take it all at once. You’ll lose money ev­ery time the fair prob­a­bil­ity changes, but that’s why they call it a sub­sidy, an this en­courages peo­ple to show their in­for­ma­tion quickly and effi­ciently.

There are ways to make that smarter, so you can lose less (or make more!) money while offer­ing bet­ter liquidity, which will be left as an ex­er­cise to the reader. Gen­er­ally they sac­ri­fice sim­plic­ity and trans­parency in or­der to make the sub­sidy ‘more effi­cient.’ The dan­ger is that if the sub­sidy is at­tempt­ing to en­sure a sucker is at the table, it does not do that if it stops be­ing the sucker, or it be­comes too hard to tell if it is one or not.

Then again, the dream is to offer a sub­sidy that doesn’t cost you any­thing, or even makes you money! Mar­ket mak­ing can be highly prof­itable when done skil­lfully, while also build­ing up a mar­ket­place.

Op­tion 2: Take Liquidity

If you provide liquidity, oth­ers will take ad­van­tage, but in some ways you make it harder to provide liquidity. If you take liquidity, you make it more prof­itable to provide it, at the risk of mak­ing the mar­ket look less liquid.

It also loses money. The more clear you are about what you are up to, the bet­ter.

There are a few fun var­i­ants of this, if you’re all right with the ex­pense.

One strat­egy is to take pe­ri­od­i­cally liquidity in both di­rec­tions. At ei­ther fixed or ran­dom in­ter­vals, ex­am­ine the or­der books in the mar­ket. If they meet re­quired con­di­tions (e.g. there is at least $X on the bid and offer within Y% of each other) then you hit the bid and lift the offer for $Z.

This costs you money, since your trades net out at a loss. If some­one else was both the best bid and best offer, they made money.

That’s the idea. You’re di­rectly sub­si­diz­ing peo­ple to ag­gres­sively provide liquidity.

Traders com­pete to be on the bid and offer to trade with you, the vir­tual cus­tomer, which in turn gives those with an opinion a liquid mar­ket to trade against. Some­times peo­ple get far too ag­gres­sive pro­vid­ing in such situ­a­tions, and those try­ing to cap­ture the sub­sidy end up los­ing money be­cause they make bad trades against oth­ers, es­pe­cially if they don’t then hedge.

You can also do this in a more ran­dom or un­bal­anced fash­ion. If you flip a coin each day and de­cide whether to be a buyer or a sel­ler, that will cause the price to tem­porar­ily be­come ‘un­fair’ to satisfy your de­mand – you’ll get a bad price. But that cre­ates a trad­ing op­por­tu­nity for oth­ers. It can also make the re­sults hard to in­ter­pret, which is a risk.

Op­tion 3: Sub­si­dize Trad­ing /​ Give Free Money

Often you’ll see crypto ex­changes do this as a pro­mo­tion, offer­ing a prize to who­ever trades the most of some coin. By pay­ing for trades, you’re en­courag­ing ex­actly what you want.

Ex­cept that you’re prob­a­bly not do­ing that. Re­mem­ber Good­hart’s Law.

The prob­lem is ‘wash’ trad­ing, where peo­ple trade with each other or them­selves with­out tak­ing on po­si­tions. This is bad on ev­ery level. It mis­leads ev­ery­one about the vol­ume and price, and doesn’t help at all with find­ing out the an­swer to the ques­tion the mar­ket is try­ing to an­swer. The last thing you want to do is en­courage it!

For that rea­son, sub­si­diz­ing trad­ing it­self is a dan­ger­ous game. But it can be done, if you’re care­ful with the de­sign.

Many on­line sites have tried this in the form of the clas­sic ‘de­posit bonus’ or even the free play. Any­one can sign up and get Free Money in ex­change for en­gag­ing in a min­i­mum amount of trad­ing ac­tivity. And of course, most of the time, a de­posit to match, if the offer is more than a small ‘free play.’ In for-profit mar­kets the goal is to have the re­quired ac­tivity make up for the sub­sidy, then hope­fully hook the cus­tomer to keep them trad­ing. There are always those look­ing to game these offer­ings if you leave them vuln­er­a­ble.

That can work for you. Get­ting those same peo­ple, who are of­ten quite cre­ative and clever, think­ing about how to come out ahead in your sys­tem can be a big win if your end goal isn’t profit! So long as you make it suffi­ciently difficult to do wash trad­ing or sign up for tons of copies of the bonuses, you can give them a puz­zle worth max­i­miz­ing (from their per­spec­tive) and effec­tively rent their la­bor to see what they think of the situ­a­tion.

Op­tion 4: Sub­si­dize Mar­ket Making

You can also sub­si­dize mar­ket mak­ing ac­tivity, as an al­ter­na­tive to do­ing the job your­self and butcher­ing it. That’s ac­tivity you can’t fake, pro­vided you set the rules care­fully. Pay­ing peo­ple who provide rather than take liquidity is good, and of­ten pay­ing for real two-sided mar­ket mak­ing ac­tivity is bet­ter. As always, make sure you’re not vuln­er­a­ble to wash trad­ing or other forms of col­lu­sion.

Op­tion 5: Advertising

Peo­ple can’t trade what they aren’t think­ing about or don’t know about.

Put­ting It All Together

Which of these strate­gies is most effi­cient and what cir­cum­stances change that an­swer?

It’s ex­pen­sive to change or clar­ify your rules and con­di­tions once trad­ing has be­gun, so in­vest in do­ing that first. Other qual­ity of life im­prove­ments are great, but take a back seat to es­tab­lish­ing good liquidity.

I list Op­tion 0 first be­cause it’s things you definitely should do if you’re tak­ing the op­er­a­tion se­ri­ously, but that doesn’t mean you always do all of them first be­fore the di­rect sub­sidy. It’s great if you can, but of­ten you need to es­tab­lish liquidity first.

If ‘no liquidity’ is the pain point and bad ex­pe­rience, there isn’t much that will over­come that. There’s no mar­ket. So if you don’t have liquidity yet, pro­vid­ing at least a rea­son­able amount, or pay­ing some­one else to do it, is the best thing you can do. Just throw some­thing out there and see what hap­pens. This makes in­tu­itive sense all around – as an easy in­tu­ition pump, if you want to know if some­thing is more likely than not, offer­ing some­one a 5050 bet on it is a great way to get their real opinion.

Once liquidity isn’t a full deal breaker, it’s time to go with Op­tion 0, then re­turn to in­creas­ing the sub­sidy and spread­ing the word.

What form should the di­rect sub­sidy take?

I’d ad­vise to con­tinue to take away bad ex­pe­riences and bar­ri­ers first.

The best sub­sidy is pay­ing to pro­duce re­li­able, safe and easy to use soft­ware, get­ting iron­clad rules in place, be­ing ready to han­dle de­posits, with­draws, eval­u­a­tion of re­sults and other has­sles. Make sure peo­ple can find your mar­kets and set up the mar­kets peo­ple want to find.

Next best is to avoid fees. Peo­ple hate fees more than they love sub­sidies. Yes, you can trick peo­ple with de­posit bonuses and then charge them a lot on their trades, but the best way to get away with that is bake the fees into the trade prices, so it doesn’t look like a fee.

At a min­i­mum, you shouldn’t be charg­ing fees for de­posits or with­draws, or for pro­vid­ing liquidity in the mar­ket.

Next up, make trades cost net zero fees. Either charge noth­ing to provide or to take liquidity, or charge a fee to take liquidity but pay it to those who provide.

After that, my opinions are less con­fi­dent, but here’s my best guess.

If that’s still not good enough, provide liquidity. Either pay some­one else to be a mar­ket maker, or provide the ser­vice your­self. I like the idea of a ‘dumb’ mar­ket maker ev­ery­one knows is dumb, and that op­er­ates with known rules that ham­string it. If you’re look­ing to provide a sub­sidy, this is a great way to do that. A smarter mar­ket maker is cheaper, and can provide bet­ter liquidity, but is less ob­vi­ously a tar­get. As the mar­ket ma­tures, you’ll want to tran­si­tion to some­thing smarter. Thin mar­kets want ob­vi­ously dumb providers.

Once you’ve done a healthy amount of that, then you’ll want to give away Free Money. Give peo­ple some cash in ex­change for par­ti­ci­pat­ing in the mar­ket at all, or trad­ing a min­i­mum amount. Or give peo­ple bonuses on de­posited funds so long as they use them to trade, or similar.

You have to watch for abuse. If you can re­spond to abuse by chang­ing the sys­tem, it’s fine to be vuln­er­a­ble to abuse in the­ory, and even al­low small amounts of it. If you’re go­ing to re­lease a cryp­to­graphic pro­to­col you can’t al­ter, you’ll need to be game the­o­ret­i­cally ro­bust, so this won’t be an op­tion, and you’ll have to re­treat to tak­ing liquidity.

Tak­ing liquidity seems less likely to mo­ti­vate the av­er­age po­ten­tial par­ti­ci­pant, and costs you weird­ness points, but does provide a strong in­cen­tive for the right type of trader. The best rea­son I can think of to use such a strat­egy is that it is ro­bust to abuse. That’s a big game if you can’t re­spond dy­nam­i­cally to un­friendly play­ers.

At the end of the day, your biggest bar­ri­ers are that peo­ple’s at­ten­tion is limited, com­plex­ity is bad, op­por­tu­nity cost is high and peo­ple don’t do things. I keep mean­ing to get around to both­er­ing with Hyper­Mind and/​or Pre­dic­tIt, and keep not do­ing it, and I’m guess­ing I am far from alone in that. Sub­sidy can get peo­ple ex­cited and make mar­kets work that wouldn’t oth­er­wise get off the ground. What I think they can’t do at rea­son­able cost is fix fun­da­men­tal prob­lems. If you don’t have a great product be­hind the sub­sidy, it’s go­ing to be or­ders of mag­ni­tude more ex­pen­sive to mo­ti­vate par­ti­ci­pa­tion.