Wide Bid-Ask Spreads in Thin Prediction Markets

# Non-arbitrageable inefficency in betting market


# Non-arbitrageable inefficency in betting market












On Polymarket, there's a market for GDP growth in 2026. There are six mutually exclusive intervals (It states that a value exactly on the border will go to the higher interval.). The odds for these intervals sum to a number much higher than $100\%$. At time of writing $153\%$, and it was $180\%$ earlier today. The obvious thing to do is of course to simply buy the "no" on each interval until the probability sums to $100\%$. The problem is that the "no" costs close to a full dollar on all of them — explaining why that hasn't already been done. In addition to the probabilities calculated by Polymarket summing to more than $100\%$, the prices of the "yes" and "no" shares for each interval generally sum to more than a dollar. In spite of all this, this is obviously normal — for some reason - because everything-ish is, due to the efficient market hypothesis, and also because, as alluded to above, it's trending toward sanity. I presume this is some sort of getting started quirk either of this type of market or of markets in general. What's going on?

Edit: I just realized, looking at the graph, that this market has existed for months (I thought maybe it didn't get created at the beginning of the year, in spite of the subject matter.). What on earth????!!!!

## Answer by Chris Taylor (score 6, accepted)

https://quant.stackexchange.com/a/85648

This is an example of the bid-ask spread.

The bid–ask spread is the difference between the prices quoted for an immediate sale (ask) and an immediate purchase (bid) in a market, such as the stock market, futures market, or a prediction market. If the bid-ask spread is wide (as it is in your example) then the market is illiquid which means you will likely see very little trading volume in the market.

## Answer by dm63 (score 3)

https://quant.stackexchange.com/a/85654

To clear up any confusion here. In Polymarket, there is no ‘house’. All prices posted are entered by participants hoping that someone will transact at their posted price. If there is a trade , Polymarket takes a cut. Ultimately, this is a simple zero-sum game between participants - the only certainty is that Pokymarket makes money.

The market displayed shows prices for 2026 GDP, but there is very little volume so the markets are wide, because not many participants seem to care about this market. There is no way to arbitrage the prices being posted.

As an example , consider the market for the interval 1-1.5%. If you want to buy Yes, it will cost you 22.8% (meaning , you stake 22.8 dollars to win 100 dollars if GDP is in that range). If you want to bet that GDP will not be in that range, you buy No for 88.7%, but you can think of this as selling Yes for 11.3%. (It’s the same economically, but Polymarket wants you to buy No because they have no credit risk to you in that style). So the two way market for Yes is (11.3%- 22.8%), which is very wide. Polymarket displays the average of these , which is 17%, on the left, to give you where ‘mid market ‘ is. If you wanted to trade in this market , the best procedure would be to improve the bid or the offer and hope someone trades with you. Crossing the bid-offer is exorbitantly expensive.

## Answer by Questor (score -2)

https://quant.stackexchange.com/a/85649

1st you need to understand. this isn't a real market. This is gambling, that is thinly veiled as a betting "market" for legal reasons (and so they don't have to abide by pesky laws that limit their overround).

### How does this 'market' ie gambling site actually work?

1st your payout is based on the odds listed. The lower the odds the more you win. And the money that you get is covered by the people who lost the bet. With enough people betting the probability that everyone picks the wining bucket is extremely low. Reducing the risk for the bookie.

If you have balanced odds (all the percentages add up to 100%) this means that the eventual payout (which is based on the odds) ensures that the bettors will take all the money that is bet. (Wait what?)

#### Lets keep it simple with fair odds.

Lets say Horse A has 2.0 odds (50% chance to win) and horse B has 2.0 odds (50% chance). someone bets 100 on Horse A, and 100 on horse B. The bookmaker collects \$200 and has to pay out 200 if either horse wins. This makes them $0 dollars in net profits... Sure people might all bet on horse A or horse B... but law of large numbers says that over time the bookie will make \$0 in net profit.

Now what happens if we push our odds past 100%.

#### Again, a simplified example:

Lets say that A/B are given odds of 1.5 instead. meaning each is given a 66.7% chance of winning, this pushes the total percentage win chance on the books to to 133% (which is insane btw as most bookies operate on a 108% margin overround).

Someone bets 100 on Horse a and Horse B. When they win they only get back 150 dollars instead of 200 dollars. And the bookmaker pockets 50 dollars. Or 33 cents for every dollar they payout. As long as the bookie keeps hthings this way law of large numbers says that they will eventually average out at making 33 cents for every dollar they payout.

So in this situation with the house pushing probabilities passed 150%.. this means that the houses payouts are low enough that they will be netting $50 in profit for every $100 that they payout (assuming enough people place bets over a long enough period of time).

TLDR: the probabilities don't work the way you think they do, the house makes money because these percentages add up the way that they do.