Searching for the "golden key" to predicting the market through 27.73 million transaction data points, yet 690 candlestick strategies still struggle to generate profits.

Author: Frank, PANews

How difficult is it to find a "golden key" to profitability in the prediction market?

On social media, you often see people claiming to have discovered a secret to making money with smart money, but these claims are often empty rhetoric. What people see are merely the profit curves of these smart money investments, not the underlying logic.

How exactly do you build a personalized trading strategy suitable for predicting markets?

PANews analyzed nearly 27.73 million transactions and 3,082 windows in the past month using BTC 15-minute prediction market data as an example, arriving at some conclusions that may challenge conventional wisdom. In a previous article , we analyzed the macro data of this market. This time, we will delve deeper to find the potential "golden key."

Illusion Shattered: The Complete Failure of Candlestick Technical Analysis

Have you ever considered a strategy that treats market prediction like trading in stocks or cryptocurrencies? By simply analyzing different entry and exit prices, and combining them with position management, stop-loss and take-profit, and other factors, you could create a trading strategy that is completely detached from BTC market movements and only considers changes in market price.

In the traditional crypto market, this trading approach is known as the "technical analysis" school of thought. Logically, this theory should also work in prediction markets. Therefore, PANews conducted simulations in this direction and developed its own prediction market backtesting simulation system. This system can calculate the actual profit/loss ratio and win rate of this strategy from over 3,000 markets over the past 30 days by inputting various factors such as entry point, take-profit point, stop-loss point, entry timing, and prices excluding interference.

Initially, with incomplete data (Polymarket only provides 3,500 historical data entries per order book), the backtesting results could easily find the answer to profitability, such as entering at 60% of the price, selling at 90% of the price, setting a stop loss at 40% of the price, and setting a trading window.

However, the actual test results were quite different. Under the actual implementation of this strategy, the profit curve showed a slow, agonizing decline. Therefore, we tried to supplement the data as much as possible. After trying various methods, we finally achieved price information data for all order books. And this time, the results finally began to match reality.

In real-world testing, PANews simulated 690 combinations of factors including price, stop-loss and take-profit levels, entry timing, interference elimination, and slippage. The final result was that no strategy could achieve a positive expected return.

Even with the highest possible return, the expected return is -26.8. This result illustrates that in the prediction market, purely mathematical predictions that exclude the events themselves are almost impossible to profit from.

For example, the "end-of-day strategy," which is widely discussed on social media, involves buying 90% of the time and selling 99% of the time. This strategy seems to have a very high win rate and should be profitable in the long run. Real-world testing shows that this strategy does indeed have a high win rate of 90.1%, achieving profit-taking 2558 times in 3047 simulations. However, the worrying aspect is that the actual profit/loss ratio under this strategy is only 0.08, and the expected return according to the Kelly Criterion is -32.2%, making it not worthwhile to adopt.

Some might argue that adding a stop-loss order would improve the profit/loss ratio. However, the harsh reality is that while the profit/loss ratio increases, the win rate will still decrease accordingly. For example, setting the stop-loss at 40% will reduce the win rate to 84%. Combined with the still low profit/loss ratio, the final Kelly expectation is -37.8%, resulting in a loss.

The most likely profitable strategy is to buy the reversal, placing a 1% bet that the market will reverse and result in a win. In simulations, this strategy has a win rate of approximately 1.1%, higher than the price probability, and a very high profit/loss ratio of 94, ultimately achieving an expected return of 0.0004. However, this assumes no slippage or transaction fees; once transaction fees are factored in, the expected return instantly becomes negative.

In conclusion, our research in this area has shown that in predicting markets, relying solely on technical analysis in financial trading is insufficient to generate profits.

The Trap of "Two-Way Arbitrage"

So, besides this approach, another mainstream viewpoint is two-way arbitrage, which means that as long as the total cost of YES+NO is less than 1, this outcome will always result in a profit. However, this is also an idea that is idealistic but unrealistic.

First, if a cross-platform arbitrage strategy is adopted, there are already a large number of bots available. Ordinary users simply cannot compete with bots for the meager liquidity available.

To achieve this arbitrage effect, another approach is to buy when both YES and NO prices drop to 40% in the same market, which would also create a 20% arbitrage opportunity.

However, the final data results were different. According to the data, although this strategy can achieve a 64.3% win rate, the low profit-loss ratio still resulted in a negative expected value for this strategy.

This "two-way strategy" seems appealing, but it's actually quite prone to failure. Furthermore, from a classification perspective, this strategy also falls under the category of purely theoretical settings detached from the actual changes in the events themselves.

Fair value and deviation models are the "golden key".

So, what kind of strategy can truly achieve profitability?

The answer lies in the "time difference" between the spot price of BTC and the price of the predicted token.

PANews found that the algorithms of liquidity providers and market makers in prediction markets are not perfect. When BTC experiences a sharp shift in a short period of time (e.g., within 1-3 minutes), such as a sudden price jump of more than $150 or $200, the token price in the prediction market does not instantly "teleport" to the theoretical price.

Data shows that the "efficiency gap" of this pricing decreases from its maximum value (about 0.10) to half (about 0.05), which takes an average of about 30 seconds.

Thirty seconds may seem like an eternity for high-frequency traders, but for manual traders, it's a fleeting "golden window."

This means that the prediction market is not a perfectly efficient market. It's more like a slow-reacting behemoth; when BTC has already set the tone, it often takes a beat longer to turn around.

However, this doesn't mean that speed alone guarantees profits. Our data further shows that the space for such "delayed arbitrage" is rapidly shrinking. In the small fluctuation range of less than $50 in BTC, after deducting gas fees and slippage, most so-called "arbitrage opportunities" are actually negative expectation traps.

In addition to momentum trading that relies on speed, PANews' research also reveals another profit logic based on "value investing".

In prediction markets, "price" is not equal to "value." To quantify this, PANews has built a "Fair Value Model" based on 920,000 historical snapshots. This model does not rely on market sentiment but calculates the theoretical probability of winning for the current token based on BTC's current volatility and the remaining time until delivery.

By comparing theoretical fair value with actual market price, we discovered the nonlinear characteristics of the pricing efficiency of the prediction market.

1. The Magic of Time

Many retail investors intuitively believe that prices should revert linearly over time. However, data shows that the convergence of determinism is accelerating.

For example, under the same BTC volatility conditions, the pricing correction speed in the last 3-5 minutes of a match is much faster than in the first 5 minutes. However, the market often underestimates this convergence speed, resulting in token prices frequently falling significantly below their fair value in the latter part of the match (the remaining 7-10 minutes).

2. Only buy at a "deep discount".

This is the most important risk control conclusion presented in this study.

Backtesting at different deviation index (fair value - actual price) levels revealed the following:

When the market price is higher than the fair value (i.e., buying at a premium), the long-term expected value (EV) is negative across the board, regardless of the price movement of BTC.

A trade has a robust positive mathematical expectation only when the deviation index is > 0.10, meaning the actual price is at least 10 cents lower than the fair value.

This means that for smart money, a price of $0.70 doesn't mean "a 70% chance of winning"; it's just a quote. Only when the model calculates an actual win rate as high as 85% is $0.70 a worthwhile "bargain" to bet on.

This also explains why many retail investors are prone to losing money in market predictions, because the actual transaction price is likely to be a purchase at a level higher than the fair market price.

For ordinary participants, this survey serves as both a sobering warning and a guide for those seeking advanced knowledge. It tells us:

Abandon the superstition of candlestick charts: Don't try to find patterns in the price charts of tokens; that's a mirage.

Focus on the underlying asset: Watch for unusual movements in BTC, rather than focusing on predicting market movements.

Respect the odds: Even with a 90% win rate, if the price is too high (premium), it's a deal destined to lose money.

In this algorithm-driven jungle, if ordinary retail investors cannot establish a mathematical coordinate system for "fair value" and lack the technical ability to capture "30-second lag," then every click of "Buy" may just be a donation to the liquidity pool.