Author: Changan, Biteye Content Team
The previous article used five methods to predict the highest temperature, but the model will not always give a unique solution: ECMWF calculates 14℃, WC calculates 13℃, and the real-time correction gives 13.5℃. Which one would you bet on?
No matter how sophisticated the weather system is, forecasts will always be limited to a range.
Moreover, the weather market carries unpredictable risks: data sources may not match, rules may change quietly, and market trends may reverse in the last hour.
Therefore, predictions must be combined with trading strategies. After two weeks of practical experience, some made money and some made mistakes, which I will share together.
Besides predictions, there are three other pitfalls.
The weather system in the previous article solved the prediction problem, but once you enter the market, you realize that sometimes the reason for losing money has nothing to do with prediction at all.
Data source mismatches, subtle rule changes, and last-hour market reversals—after experiencing these pitfalls, I realized that the risks in the weather market aren't limited to forecasting; there are three layers in total:
1. Data source issue: WU and METAR do not match.
The rules in the weather market generally state that WU data is the standard.
WU stands for Weather Underground, a US-based weather platform whose data comes directly from observational records reported by weather stations around the world. For the weather market, WU accesses data from local airport weather stations.
Every half hour, the airport weather station sends a standardized weather report called METAR, a format used globally by civil aviation. It contains information such as temperature, wind speed, cloud cover, and visibility, and is a crucial basis for air traffic control. Theoretically, the airport temperature displayed on the WU should be derived from this METAR report.
This is indeed the case at airports in other cities; WU readings and METAR readings are largely consistent, with negligible errors. Many traders have therefore developed the habit of directly monitoring METAR, treating it as a real-time preview of the settlement temperature.
However, the WU data and METAR data at Shenzhen Bao'an Airport often do not match, and there have even been cases where the difference is as much as 2 degrees.
This bias is not a concern in other cities, but in Shenzhen, it can turn a perfectly correct transaction into a loss.
2. Frequent rule changes
Perhaps because WU and METAR had been mismatched for a long time, Polymarket changed its data source for the Shenzhen market on March 29, switching the settlement data source from WU to NOAA.
The rules page has an update record dated March 28th, containing only one sentence: "This market's resolution source has been updated."
NOAA uses data from weather.gov, which is closely aligned with METAR, but often deviates from WU's readings.
There's an account called ilovebigbiscuit that bet "No" on the 27°C level, averaging 99.8¢, resulting in a loss of $7,883. Most likely, the WU data showed the temperature wouldn't reach 27°C, leading them to believe it was safe, only to have the NOAA readings contradict them, causing them to lose everything by trading at the close.
The Shenzhen weather market later switched back to WU. One market changed its data source twice within just a few days.
Therefore, make it a habit: every time you enter a new weather app, the first thing you should do is not check the temperature, but click on the Rules in the lower left corner to confirm which data source is being used for settlement. Skipping this step may render any subsequent analysis futile.
3. Continuous reversal trend
In the past two weeks, the Shanghai weather market has repeatedly shown the same trend: a certain temperature range leads the way from the morning and steadily suppresses other ranges. Just as the settlement is about to be completed, the trend suddenly reverses in the last hour, with another range jumping from close to 10% to 100%.
As shown in the image, 20°C was the dominant level from the morning, rising to nearly 90% by 2 PM, seemingly a done deal. However, after 3 PM, 21°C reversed from almost 0% to 100%, ultimately settling at 21°C. Those who bet on 20°C were correct until the last hour, but were all wrong at the settlement.
Spring weather in Shanghai is inherently unstable, and afternoon temperature trends are greatly affected by cloud cover and wind speed, so any predictions made in the morning may become completely invalid in the afternoon.
II. Four Practical Fighting Methods Observed Over Two Weeks
Betting on a single temperature level is too difficult to predict correctly, so most players buy several adjacent price levels simultaneously. As long as the total cost of these price levels doesn't exceed $1, it's profitable. However, even with multiple temperature levels, the timing, method, and market choice can lead to vastly different results. Here are some trading strategies observed over the past two weeks:
Strategy 1: Buy low-priced shares a few days in advance
Another completely different approach utilizes the inherent uncertainty of the weather.
Polymarket's weather market opens four days in advance. The earlier the market opens, the more dispersed the pricing across different price ranges becomes, and many temperature probabilities are not yet fully priced in, remaining below 5¢.
Some players specifically exploit this time difference. Today is April 1st, so they go to the weather market on the 4th and scan all the price tiers below 5¢, buying as long as it's cheap enough. The logic is simple: the forecast is for three days, and the weather could change at any time. Temperatures that seem impossible today might become hot tiers in a couple of days, and 5¢ chips might rise to 30¢, 50¢, or even higher.
The core of this strategy is betting on a change in the market trend. The position is held until the end of the day, and as long as enough price levels are covered, the total cost is kept below $1. The final settlement price level recovers $1, while the rest become zero, resulting in no overall loss. However, if a price level experiences a significant upward movement, profits can be locked in by selling earlier.
Strategy 2: Use meteorological factors to capture underestimated temperatures
Popular tiers are usually fully priced in, making them expensive to buy into and offering very low odds. However, there's a group of traders who do the opposite, specifically looking for undervalued, less popular tiers.
These traders monitor real-time weather factors. For example, if it's 1 p.m., they'll look at the wind direction and speed for the next hour or two: southerly winds usually bring warm, humid air and have the potential to raise temperatures; if wind speed, cloud cover, air pressure, and other factors all point in the same direction, they'll bet on the temperature level that the market hasn't yet reacted to.
Because it's a less popular price range, the price is low, and the entry cost is small. Even if the judgment is wrong, the loss will be limited. However, if the judgment is correct, the price increase of low-priced shares can be very considerable.
WU's data is updated every half hour. If the latest data shows that the meteorological factors are not developing as expected, such as the wind direction changing or the temperature rise stopping, then sell directly to stop the loss.
This strategy requires a high level of meteorological knowledge and a true understanding of how factors such as wind direction and cloud cover affect temperature. It cannot be judged by simply glancing at a forecast. It is suitable for traders with a professional background or those who have been working in this market for some time.
Strategy 3: Closing Strategy
There's a pattern in Shanghai's weather market: temperatures generally stop rising after 3 PM, and the highest temperature usually occurs before that.
The closing strategy utilizes this window. WU's data is updated every half hour. After the price surge stops, monitor WU's half-hourly updates and enter the market the instant the data is refreshed. At this point, Polymarket's price hasn't had time to react, and you usually only gain a few points.
There are two possible strategies: buy the "Yes" option at the current temperature, or buy the "No" option at the next higher temperature. These two strategies are essentially the same, because the temperature increase has stopped, leaving only these two possibilities. Which one to choose depends on which price is more advantageous.
The biggest risk of this strategy is the unpredictable nature of the warming period. The time when the warming stops each day is not fixed; cloud cover, wind speed, and air masses all affect the pace of warming that day. If you judge that the warming has ended, but the temperature subsequently rises again, your judgment at the end of the day will be completely invalid.
Strategy 4: New Market Order Placement Strategy
Polymarket's New Weather Market has a notable characteristic: there are no market makers, and the spread is very large. The difference between the best bid and best ask prices can be tens of cents, for example, a best bid of 20¢ and a best ask of 60¢.
This spread is the opportunity.
The specific operation involves placing buy orders at popular temperature ranges and waiting for them to be filled. Because the spread is large enough, even if buy orders are placed at several popular ranges, the total cost can be kept below $1.
However, these types of markets have one crucial characteristic to note: extremely low liquidity. A transaction of just a few hundred dollars can drastically reduce the probability of the current hottest trend, making it appear as if it's about to lose.
Therefore, there is only one principle to follow when implementing this strategy: once the order is filled, hold it until settlement and avoid frequent trading. Short-term price fluctuations are not relevant in a low-liquidity market.
Strategy 5: Establish a position one day in advance (counter-example)
Wu will release the high temperature forecast for the next day one day in advance. The most intuitive way to do this is to refer to this forecast and buy in the three price ranges around the predicted temperature one day in advance.
One player did just that. On March 27th, a day in advance, they bought three temperature settings: 15°C, 16°C, and 17°C. On March 28th, they bought three more temperature settings: 19°C, 20°C, and 22°C. They invested several hundred dollars in each setting, keeping the total cost under $1.
However, weather forecasts are not fixed; WU's predictions are adjusted in real time as the weather changes. Today's forecast is 22°C tomorrow, but by tomorrow morning it might have changed to 19°C. If you open a position a day in advance, locking in last night's forecast, the temperature will have already deviated significantly by the time you settle the position.
All results were zeroed out. The temperature was 19°C on the 27th and 21°C on the 28th, with the covered range differing from the actual temperature by two or three degrees.
The idea of multi-temperature coverage is correct, but the entry time is too early. Establishing a position before the forecast has stabilized is equivalent to betting on today's outcome with yesterday's information.
Strategy Six: The Winning Rate Trap of Buying No (Counterexample)
Some commenters said that betting on "Yes" is too difficult to predict, and they might not win at all, so betting on "No" has a higher win rate. But is that really the case?
Weather markets typically have 11 temperature ranges. Buying "Yes" means guessing one out of 11 correctly, while buying "No" means guessing 10 out of 11 correctly. "No" seems to have a natural advantage in terms of win rate, which makes sense.
However, the No. 1s in popular price tiers are usually priced above 80¢, and only a few tiers have such prices. If we buy all the No. 1s above 80¢, assuming there are 4 popular tiers:
Cost: 4 × 80¢ = $3.20
The settlement will most likely fall into one of these four popular tiers, meaning 3 No wins and 1 No loss:
Win: 3 × 20¢ = 60¢
Loss: 1 × 80¢ = 80¢
Net result: Loss of 20¢
The win rate is indeed high, but the odds completely negate this advantage. Each win only yields a small profit, while a single wrong guess results in a loss that outweighs several previous wins. The price of "No" already incorporates the win rate, so betting on "No" offers no additional advantage.
III. Forecasting and strategy are both indispensable.
The forecasting system is your "eyes," and the trading strategy is your "armor." Together, they create a complete and invincible force in weather market trading.
The weather market is still in its early stages; the rules are unstable, data sources change, and market trends can reverse repeatedly. But precisely because of this, information asymmetry exists, and so do opportunities.
If you're also working in the weather marketplace, feel free to share your operational strategies and troubleshooting experiences in the comments section. 👏
