What Polymarket's Israeli election markets say, and who is behind the prices

What the data show

1. The markets now favour Gadi Eizenkot and Yashar. Eizenkot is at 51¢ to be the next prime minister, Benjamin Netanyahu at 27¢, Naftali Bennett at 13¢ and Avigdor Lieberman at 3¢. For largest party, Yashar is at 72¢ and Likud at 28¢. Prices have swung: Likud briefly drew level in early October (Yashar 49¢, Likud 51¢ a week ago) before Yashar regained and widened its September lead, and the chance of Likud winning fewer than 20 seats jumped from 14¢ to 40¢.
2. A week out, election prices have been well calibrated. In 8,631 resolved election markets, contracts priced near today's Israel prices 6–10 days before the event won at about their price (section 2): 27% for contracts near 27¢, 48% near 51¢, 14% near 13¢ and 73% near 72¢. On that record the prices can be read as probabilities, with an error of a few points either way.
3. The experienced money is positioned against Netanyahu; the new money is for him. Grouping wallets by their record in every other market, proven forecasters, wallets whose past large orders beat the price, and market makers would all lose if Netanyahu wins and gain if Eizenkot wins. Wallets with a short record, or none outside Israel, hold the mirror-image book: together they gain about $1.4m if Netanyahu wins. The gradient is steep: wallets that made more than $100k elsewhere are positioned $1.6m toward Eizenkot over Netanyahu, and only 28% of them lean Netanyahu, against 66% of wallets with no record outside Israel (figure). Likud's seat market tells the same story: experienced books profit if Likud wins fewer than 25 seats, while short-record and no-record wallets are on 30–34. The largest-party market shows a weaker version: wallets that made over $10k elsewhere, proven forecasters and market makers lean Yashar and wallets with no record outside Israel lean Likud, but the broad "profitable, no proven edge" group leans Likud. It is not a by-product of Netanyahu's falling price: in the 11 weeks his price rose, experienced wallets still sold him in 10 (figure). In past markets, the side experienced money traded this steadily won about as often as its price implied, not more (figure). This says who moved the price, not that it will move further: their selling is a large part of why Netanyahu is at 27¢ rather than 35¢. Across 27,000 markets a week before resolution, knowing which side experienced wallets hold did not improve on the price (traders page, section 7), and here proven forecasters' remaining net tilt against Netanyahu is under 4% of all positions.
4. Large buys of Netanyahu have mostly reversed. Over the last month, many of the biggest orders in the Netanyahu market were purchases of $15k–$100k at 29¢–37¢, about 60% of them by wallets with a short record or none outside Israel. The price has since fallen to 27¢. In the full universe, orders from such wallets carried no information on average and their price impact faded (traders page), which is what happened here (section 5).
5. Much of the volume in long-shot candidates is not a view. Markets on candidates priced under 1¢ (Barkat, Katz, Golan, Sa'ar, Cohen) each show $290k–$760k of trading among 900–3,400 wallets, and of the trades that can be matched to a counterparty, 53–92% are between just five pairs of wallets. That is the signature of wash trading or reward farming, not information (section 6). In the three leading candidates' markets, the five busiest pairs account for only 8–20%.
6. No sign that locals know more. Wallets whose trading hours place them on Israel's clock are not over-represented once two wash-trading wallets are set aside, earned no more in past Israel-related markets than elsewhere, and lean the same way as everyone else today (section 8).
7. No sign of a campaign to hold the price up. In the main candidate markets, the mix of large orders that moved the price and then faded, or were pushed again after fading, is about what other election markets of the same size show (9% of large-order dollars, against 9%). No wallet pushed the same Israel market again after a fade more than twice. The small amount of re-pushing ($119k) is mostly purchases of Netanyahu by wallets with a short record; in past election markets of this size, such re-pushed orders ended up on the losing side by about 6¢ (section 7).

1 · The markets now

The prime minister market is one market per candidate; each Yes share pays $1 if that candidate becomes prime minister, so the prices across candidates add up to about 100¢. The largest-party market works the same way. Over the spring and summer, Eizenkot rose from about 10–15¢ to about 50¢, largely at Bennett's expense (Bennett fell from 37¢ in May to 6–15¢). Netanyahu drifted down from about 40¢ to 27–35¢, with a rise to 35–37¢ in the first ten days of October that has since reversed.

2 · How far to trust prices a week out

The companion page on calibration shows that Polymarket prices are close to calibrated at short horizons, and that election markets are the best informed category a week out (Brier skill about 0.6). Applied to the Israel contracts, the historical record of election markets at the same horizon and similar prices says the current prices are about right on average. That does not say whether this election is an exception; it says that, historically, a 51¢ election contract a week out has won about half the time.

3 · Who is trading: the wallets' record in every other market

Of the 51,995 wallets that traded the Israel markets, 95% also traded at least one of the 74,000 resolved markets in our data, so their behaviour elsewhere can be measured. As a group they are more experienced and more successful than the typical Polymarket wallet: 37% made money overall, against 30% of all wallets; 83% have traded election markets elsewhere and 68% other Israel-related markets (such as Gaza, Iran or Netanyahu markets).

The groups below sort them by that record. The lessons from the full universe say which groups to weight: wallets with a proven, persistent edge and wallets whose large orders beat the price carried information; new and short-record wallets did not, and their large orders tended to reverse (skill persistence, wallet types).

4 · Where each group's money is

In a market with several candidates, a wallet's position in any single candidate is misleading: a market maker who sells "Yes" on every candidate is not betting against all of them. What a book really bets on is visible in its profit under each possible result, because exactly one candidate wins. For each group, the table shows the total profit or loss of its wallets' current positions under each outcome, including what they have already paid and received.

Two caveats. Market makers' books partly reflect the other side of what retail traders bought from them, not a view of their own. And the table shows direction, not how strongly a group disagrees with the price: a group positioned against Netanyahu thinks his chances were below the price at which it traded.

Netanyahu started the year as the favourite, at 52–64¢, and has fallen to 27¢. Is the experienced money against him simply because his price has been falling, so that traders who sold early look smart? If so, the experienced wallets' trading would track the trend: buying him in weeks he rose and selling in weeks he fell. They did not. In the 11 weeks since January in which Netanyahu's price rose, wallets that made more than $10k elsewhere were net sellers of Netanyahu against Eizenkot in 10, on average 24% of the week's volume; in the 22 weeks his price fell they sold in 17, on average 14%. They sold into his rallies even more than into his declines, and their weekly trading is uncorrelated with the price change (−0.09). Wallets with no record outside Israel did the mirror image in both kinds of week. The wallets in between are the ones that followed the trend, buying Netanyahu in weeks he rose. Across all Polymarket markets, groups' weekly trading tends to move slightly with the price (correlations of about +0.1), so the experienced wallets' steady selling here stands out.

The big fall in Netanyahu's price, from about 60¢ to 40¢ in March, came before these wallets held any position worth mentioning. Their position was built gradually, mostly from August onward as volume grew, when his price was already around 30–35¢. It reflects a consistent view that his price was too high, held through his rallies, not a reaction to his decline.

How to read this. Two questions are easy to mix up. The first is whether the experienced wallets have a steady view: they do, and it is not a reaction to the trend. The second is whether that view is more right than the price. Their selling is one of the things that set the price: Netanyahu is at 27¢ partly because they kept selling to anyone who would buy. For their position to be a signal on top of the price, markets would have to stop short of where experienced money wants them, and the side it trades would win more often than its price implies.

That can be checked in the past. In 3,617 resolved markets, the net trading of wallets that had already made more than $10k elsewhere was measured over the 8 weeks before the week-out point, as here. The side they traded toward was usually the favourite, and it won more often than not: 66% of the time in the 636 markets where they traded it consistently. But its price a week out was already 65%. In election markets the consistent side won 60% against a price of 61%. In no group of markets, however strongly or consistently the experienced money traded, did its side beat its price by a margin that could be told apart from chance. In the Netanyahu market today, the side they hold, "not Netanyahu," is priced at 73%; on this record, that is about how often it should be expected to win.

Profit = cash paid and received so far plus the payoff of the shares held, if that outcome happens. Blue cells are the group's better outcomes, orange its worse ones, relative to its own average. "Anyone else" is any other candidate or party winning.

5 · Large orders in the last 30 days

In the full universe (see traders page, section 5), about 38% of large orders are followed by momentum (the price keeps moving their way) and 35% by reversal. Reversal was more common after orders from new and short-record wallets, and momentum after orders from dominant one-sided traders. In the Netanyahu market over the last month, $574k of large orders have so far been followed by reversal and $408k by momentum. The reversals are concentrated in purchases of Netanyahu: about 60% of them by wallets with a short record or none outside Israel, the rest mostly by market makers.

6 · Trading that carries no view

Polymarket pays liquidity rewards, and an estimated quarter of its volume is wash trading: accounts trading mainly with each other (Sirolly, Ma, Kanoria and Sethi, 2025). In the candidate markets below, the share of all matched volume traded between the five most active pairs of wallets is a simple indicator. It is very high in the long-shot candidates' markets, so their volume figures say nothing about interest in those candidates.

7 · Is anyone pushing the price?

Can manipulation be told apart from honest trading? Not trade by trade: someone who wants the price rather than the payoff leaves the same trace as an overconfident bettor. They pay to move the price, the move fades, and then they push again or sell into the move. The traders page sorts all 757,000 large orders on Polymarket by this trace and checks how they turned out. Orders whose move faded were on the losing side by 4–6¢, while orders whose move held were on the right side by 8.5¢. Sustained price support, meaning the same wallet pushing the same market again after a fade three or more times, accounts for about 1% of large-order dollars.

Applied to the Israel markets, the trace gives a measure that can be compared with other markets before the result is known. Large orders in the four main candidate markets break down almost exactly like those in other election markets of the same size. A little more of their volume moved the price and held. Less was sold back within a day. The share that faded, with or without another push, is about the same. Nobody ran a sustained campaign. The orders that were pushed again after a fade total $119k, and $76k of that was purchases of Netanyahu by nine wallets; two-thirds of all re-pushed dollars came from wallets with a short record elsewhere. Those buys reversed (section 5).

This does not rule out manipulation that the trade log cannot see. Trades through several unrelated wallets, or patient buying that never moves the price by 1¢, would not show up here. Nor does it cover attempts to move the outcome itself rather than the price.

8 · Are locals trading, and do they know more?

Every trade carries a timestamp, so each wallet's daily rhythm hints at where its owner lives: the hours when it never trades are its night. That makes it possible to ask questions that stock-market data cannot answer: do people near an event trade on it, and do they know more?

Two limits matter. A wallet can be placed only if it trades often and has a clear night. About a third of wallets with at least 40 orders do (the rest trade round the clock, like bots and market makers, or too irregularly), and in the Israel markets they account for about a quarter of the dollar volume. And a time zone is not a country: Israel's clock is shared with Eastern Europe, Turkey and western Russia. An attempt to pick out Israelis by their week also failed. Wallets with an Israeli weekend (Friday and Saturday different from Sunday to Thursday) are no more common on Israel's clock than elsewhere. The same goes for wallets that never trade from Friday evening to Saturday evening: about 5% on Israel's clock, about the same as in other time zones.

With those limits, three results. First, among the wallets that can be placed, those on Israel's clock supply 31% of the Israel markets' dollar volume, twice their 16% share of all Polymarket orders. But two wallets account for most of it, and both end with no net position after trading long-shot candidates back and forth: the wash-trading pattern of section 6. Leaving out the five largest, Israel's time band supplies 7% of the volume, no more than its share of trading elsewhere. Second, there is no sign of a local edge in the past. In 1,292 resolved Israel-related markets, wallets on Israel's clock earned 1.3% per dollar (95% interval −4% to +9%), less than they earned elsewhere. Their large orders beat the price by 1.6¢ there, against 3.3¢ in other markets, and wallets in the Americas did at least as well in Israel-related markets. Third, today they are positioned like everyone else. In the prime-minister market, 40% of the placed wallets on Israel's clock lean toward Netanyahu over Eizenkot, the same as in the Americas, and over the last month they traded neither way on balance.

How this was built