Hand placing a paper ballot into a transparent ballot box

FuturesAI Election Odds: How to Score a Forecast

Short answer: A forecast price only becomes a usable signal after you score it. FuturesAI’s election article reports final 2024 prices of 61 cents (Kalshi) and 62 cents (Polymarket) against model forecasts of 50% to 56%. Scored with a Brier formula against the actual result, the markets land at 0.152 and 0.144 and the models at 0.194 to 0.250. That is one event, though, so it proves nothing about the next one.

On October 31, 2024, two contracts and three models were all looking at the same race.

Afterwards everyone said the markets won. That is easy to say and a little harder to measure.

This post scores the numbers from FuturesAI’s article with a simple formula, then shows what a trader would have needed to pay in fees to act on a price. It is arithmetic, not advice. Prediction markets can lose the full stake, and nothing here is investment advice.

Check it on FuturesAI

Where the Numbers Go

What did the 2024 forecasts say?

FuturesAI’s article on Kalshi election accuracy reports that on October 31, 2024 Kalshi’s Trump-wins-presidency contract closed at $0.61 and Polymarket’s equivalent at $0.62, while 538 gave 50%, Silver Bulletin 53% and The Economist 56%.

The same article says Trump went on to win, and that markets were the most bullish major forecast on that outcome. It also puts real money behind the prices: $9B on Polymarket and $470M on Kalshi election volume.

Hand placing a paper ballot into a transparent ballot box
FuturesAI reports final 2024 presidential prices of $0.61 on Kalshi and $0.62 on Polymarket (article read November 4, 2026).

One note on dates. The article also quotes an opening price for the 2026 midterms. The vote has now happened, so I left that number out instead of treating an old opening price as a live signal.

How do you score a forecast with a Brier score?

You score a probability forecast with a Brier score by subtracting the forecast from the actual outcome (1 if it happened, 0 if not) and squaring the result, so a lower number is better and a 50% guess always scores 0.25.

For a Yes outcome that works out to (1 minus the forecast) squared. Here are the five forecasts from the article.

Forecaster Forecast for Trump Math (outcome = 1) Brier score
Polymarket 62% (1 – 0.62)^2 0.1444
Kalshi 61% (1 – 0.61)^2 0.1521
The Economist 56% (1 – 0.56)^2 0.1936
Silver Bulletin 53% (1 – 0.53)^2 0.2209
538 50% (1 – 0.50)^2 0.2500

When I computed these, I found that the spread between the best and worst is only about 0.106. The ranking matches the headline: the more bullish the forecast, the lower the score, simply because the outcome was Yes.

Business charts, calculator and eyeglasses on a desk
A one-event Brier score ranks 62% best and 50% worst for a Yes result, which is just the ordering of the forecasts.

Why does one scored event prove so little?

A single scored event proves little because the most bullish forecast always wins the Brier ranking when the outcome is Yes, and the least bullish always wins when it is No, so the ordering reflects the result, not skill.

Flip the outcome to No and the same five forecasts score like this.

Forecaster Forecast for Trump Math (outcome = 0) Brier score
538 50% 0.50^2 0.2500
Silver Bulletin 53% 0.53^2 0.2809
The Economist 56% 0.56^2 0.3136
Kalshi 61% 0.61^2 0.3721
Polymarket 62% 0.62^2 0.3844

The order reverses completely. That is the whole point. Someone who always says 62% looks like a genius or a fool depending on a coin that landed once.

The article itself points the same way: it says markets were less accurate on individual House and Senate races and that liquidity matters, because the lower the volume, the less efficient the price. I noticed that this is the only part of the page that comes with a reason rather than a number.

What does a taker fee do to a 61 cent price?

A taker buying Yes at $0.61 in a Polymarket politics market pays a fee of $0.0095 per share under Polymarket’s fee formula, so the contract needs to be right about 62% of the time just to break even.

Polymarket’s documentation, read November 4, 2026, gives the fee as shares times fee rate times price times (1 minus price), lists 0.04 for Politics, and says makers are never charged. Applying that to a few prices, with the price and category as my own illustration:

Yes price Fee per share (0.04 x p x (1-p)) Cost per share Break-even win rate
$0.40 $0.0096 $0.4096 41.0%
$0.50 $0.0100 $0.5100 51.0%
$0.61 $0.0095 $0.6195 62.0%
$0.62 $0.0094 $0.6294 62.9%
$0.80 $0.0064 $0.8064 80.6%

The fee is symmetric around 50%, which matches Polymarket’s note that a trade at 30 cents pays the same dollar fee as one at 70 cents. At 61 cents it is about one cent a share. Small, but it moves the bar from 61% to roughly 62%.

I did not use a Kalshi fee here. I could not reach an official Kalshi fee schedule, so any Kalshi number would be a guess.

Which filters does the article suggest?

FuturesAI’s article suggests four filters for reading election prices: trust top-line high-volume markets more than long-tail ones, treat a market-versus-poll gap above 5 points as meaningful, compare against fundamentals models, and watch for Polymarket and Kalshi disagreeing by more than 3 cents.

It attaches a number to one of them: when markets and polls diverge by more than 5 points, the market is right about 70% of the time on contested races, from a 2008 to 2024 sample. The page gives no count of races. Without the count, 70% could be 7 of 10 or 70 of 100, and those are very different amounts of evidence.

Filter (from the article) What it checks What is missing
High volume only Is the price liquid enough to mean something? No volume cut-off is given
Gap over 5 points vs polls Is the market saying something the polls are not? About 70% hit rate, no race count
Gap over 3 cents across venues Does one venue have a structure issue? No test of which venue is wrong
Add a fundamentals model Is news outrunning slow fundamentals? No model named

Signal filter checklist

  1. Write down the price and read it as a probability.
  2. Check the volume behind it. Thin books are not signals.
  3. Ask for the sample size behind any hit rate before you use it.
  4. Add the taker fee and move your break-even up by that amount.
  5. Decide the maximum you can lose on the position first.

Check it on FuturesAI

Tips

Warnings

  • This is not investment advice. Prediction markets and leveraged products can lose the full amount you put in.
  • FuturesAI makes no profit promise here, and past forecast accuracy does not predict future results.
  • Fee rates and categories change. These were read on Polymarket documentation as of November 4, 2026.

FAQ

Does a lower Brier score prove a forecaster is better?

No. A Brier score from one event, such as the 2024 presidential race, only shows who was closer once. FuturesAI’s own article notes that markets were less accurate on individual House and Senate races, so a single headline result is not a track record.

What does a 61 cent Kalshi price mean?

A Yes contract priced at $0.61 pays $1 if the event happens, so the price is read as a 61% implied probability before fees and spread. FuturesAI’s article reports Kalshi’s Trump contract closed at $0.61 on October 31, 2024.

Do prediction market trades have fees?

On Polymarket, takers pay a fee of shares times the fee rate times price times one minus price, makers pay nothing, and Politics carries a 0.04 rate according to Polymarket’s fee documentation read on November 4, 2026. Geopolitical and world-events markets are fee-free.

Sources

Figures read on the FuturesAI article and Polymarket fee documentation as of November 4, 2026. The Brier scores and break-even rates are my own arithmetic on those numbers. Not investment advice.

๐Ÿ“Œ Hub guide: For the full earn-save-flip playbook โ€” pricing, taxes, cashback, and collectible margins โ€” see the Income & Smart Shopping Hub.

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *