The author previously shared several tools built through Coding in an article about creating an AI-powered investment workstation. Those tools included a cross-market asset dashboard, an investment map, a personal content operations desk and, most recently, a Polymarket betting monitoring panel that has been used frequently.
Over the past half month, the author tested the Polymarket dashboard with roughly $1,600 in principal and recorded a return of more than 30%. The real-time statistics shown by the dashboard were largely consistent with the final actual net profit, with only about 6U of difference. That gap mainly came from small deviations such as pending orders and liquidity rewards.
However, the point of the review is not to present Polymarket as an easy place to make money, nor to package the dashboard as an arbitrage tutorial. The author’s conclusion is almost the opposite: after completing this round of testing, Polymarket looks less like a venue suited for aggressive “arbitrage” and more like a market that requires information discipline, position control and careful treatment of binary risk.
From manual Excel tracking to a real-time PM dashboard
The dashboard was first built around May 21. The initial need was simple: the author did not want to open more than a dozen betting pages every time just to check how yes/no prices were moving, and did not want to keep recording trades, floating profit and loss, settlement dates and event types manually in Excel.
That manual approach had created practical problems. In Polymarket trading, a small position can easily become larger when odds move and the trader wants to add. When an event suddenly shifts and spreadsheet data has not been updated in time, the window for stop-loss action or position increase can be missed. The author describes the whole process as too fragmented; without a system, people are more likely to place orders based on emotion.
The dashboard was therefore designed to put each bet back into a unified framework, turning vague impressions into information that can be visualized and compared horizontally. After several iterations, it was divided into two tabs: a “position dashboard” and an “opportunity monitor.” The position dashboard is the core component, functioning as a dynamic system that captures PM real-time data and recalculates key metrics.
For each market, the panel records the event name with a hyperlink to the trading page, a T1/T2/T3 tier judgment, current yes/no prices, yield, annualized yield, abnormal movement alerts, observation checkpoints and a countdown to settlement. The abnormal movement threshold can be customized. For example, if the price changes by more than 20% within 24 hours, the dashboard can trigger a pop-up reminder as long as the webpage remains open.
The author highlights two design details. First, by finding a suitable PM interface, the dashboard can parse a betting event link directly and automatically show yes/no options, corresponding prices and categories of different outcomes under the same event. This greatly reduces manual input. Second, the tier classification of the same bet can automatically adjust as the remaining time changes.
Expected value is not the same as safety
Before Anthropic released Mython, the author observed clear price movement in the watchlist. In that case, the event was judged to be close to a high-probability outcome, and entering at that stage could capture roughly 10 points of return. Without a watchlist, this type of opportunity would be hard to catch consistently.
Still, the author’s broader point is the structural trap in binary markets. Such markets are unfriendly to traders who like concentrating heavily in a single position, while they are more suitable for participants who are used to running a diversified “supermarket” of many positions.
The author gives a simple expected value example. Suppose the yes price of an event is c = 0.80, which means the market is roughly pricing an 80% chance of occurrence. If the author’s own estimate of the true probability is q = 0.90, then the expected return can be roughly calculated as EV = q / c - 1 = 0.90 / 0.80 - 1 = 12.5%.
That looks attractive, but Polymarket is not a bond. Behind that 12.5% expected return sits a sharp tail risk. If the judgment is wrong, the loss is not 12.5%; it is 100%. For that reason, the dashboard does not focus only on expected return. It also tracks the gap between the author’s probability judgment and the market price, namely q - c, which is the core of whether an edge really exists. The author also sets an automatic take-profit reminder target at the midpoint between the entry price and 100.
The second metric is the impact of a single position going to zero on the entire account. This is the basis for the T1, T2 and T3 classification system. The purpose of position limits is to reserve room for the cost of being wrong.
T1, T2 and T3: conviction does not justify unlimited size
The author divides positions into three categories. T1 refers to high-conviction bets. The author’s comfort zone includes East Asia and some geopolitics-related events, especially where an information gap between Eastern and Western contexts is believed to exist and can be checked repeatedly before entry.
T2 refers to relatively stable opportunities where the current implied probability appears clearly different from the author’s assessment of the actual yes or no pricing. T3 refers to pure speculation, usually with high odds. The author does not hold these often and prefers using them for short-term counter-moves, aiming to benefit from price reversion rather than relying on them for large profits.
T1 also carries hidden costs, especially when the event has a long settlement horizon. For example, if a T1 bet has a static return of 18% but settles 180 days later, its annualized IRR may be only 3–4%. During that period, principal is locked and later high-IRR opportunities may be missed. The author therefore further divides T1 internally by time bucket: short-term T1-A can receive more allocation, while long-term T1-C requires restraint. Over-allocating to low-IRR long-term bets is an implicit loss of capital efficiency.
For T2, even when there is edge, room must be left for being wrong. The author sets a single-bet cap of 8–10%, meaning that even if the position goes to zero, the total account loss remains within 10% and does not prevent participation in later opportunities. T3 positions are kept at the smallest size for observation. The goal is not to make a large amount from them, but to keep tracking high-odds events and build a feel for this kind of market.
A counterintuitive but important point is that high conviction does not mean high position size. Even if an event is believed to have a 95% chance of occurring, the remaining 5% chance of going to zero still requires a limit on exposure. If a trader makes 10 independent bets each believed to have a 95% win rate, every individual trade looks stable. But the probability that at least one of them is wrong is about 1 - 0.95^10 ≈ 40%.
Real PM markets are often even more complicated because many of them are correlated. The author cites examples such as “whether U.S.-Iran talks reach an agreement,” “whether the Strait of Hormuz reopens,” and “whether the Middle East situation escalates within the month.” These look like three separate markets, but the underlying variable is nearly the same: the direction of Middle East geopolitical policy. If that direction is misjudged, all three positions can suffer at once.
PM as a judgment training ground, not a stable yield machine
The author says the dashboard’s greatest value is not improving win rate, but preventing major mistakes. Its core value is risk control rather than profit statistics. After more than half a month of intensive testing, the conclusion is that Polymarket does contain opportunities, but it is not the arbitrage venue many people imagine.
In previous on-chain arbitrage, rules were often clear and price dislocations could be locked in. Polymarket is different. It tests the trader’s logical understanding of how the wind can shift across different outcomes. For East Asia-related political and economic developments, Chinese-language users can have an information-gap advantage, but that does not guarantee a win.
Polymarket ultimately settles not according to “the reality as you understand it,” but according to market rules and specified data sources. The author also notes that UMA manipulation issues have appeared repeatedly. In addition, something that looks settled in a Chinese-language context may not match the definition used in English rules. Many betting rules contain textual traps.
Based on the author’s experience, PM does not offer many true arbitrage opportunities. The main approach is still information gaps and position diversification, and even high-conviction bets can meet black swan outcomes. When that happens, the principal is gone. As one friend put it, “When it comes to investing, even if there is only a 1% probability of going to zero, one should not rely on luck.”
The author’s current understanding of PM is therefore more conservative. It should not be treated as a stable yield tool. Even high-conviction positions should not create the illusion of a cash machine after several consecutive wins. Binary markets are dangerous because they can make a trader believe, after a winning streak, that every judgment is correct; then the final position becomes too large and gives back the earlier profits.
The author also stresses that a high win rate is not the same as a good trade. An event with a 90% win rate may be negative expected value if the market price is already 0.95. Conversely, an event with only a 40% win rate may have positive expected value if the market price is only 0.20. Tail risk must also not be ignored. A position offering 10% or 20% return can still be far from low risk if the wrong outcome means a complete loss.
Another warning is against pseudo-diversification. Buying multiple markets does not automatically mean risk is diversified. If several markets are driven by the same underlying variable, as in the U.S.-Iran talks, Strait of Hormuz and Middle East escalation examples, they can fail together. For this reason, the author increasingly treats PM as a training ground for judgment rather than as a direct arbitrage machine.
Private-market valuation signals and PM price movements
Besides the PM betting dashboard, the author also used Codex to build a dynamic private-market valuation monitoring panel. It mainly tracks valuation changes of unlisted unicorn companies such as Anthropic, OpenAI, Stripe and Kraken in the private market, as well as the relationship between those changes and corresponding bets on PM.
Polymarket is essentially an expectations market. Sometimes signals in the private market have already changed, while PM prices have not moved yet. At other times, PM prices move first while real-world data has not caught up. The misalignment between the two is something the author wants to continue observing.
However, the author emphasizes that this is not risk-free arbitrage either. Private-market valuations are not fully transparent, and different data sources can show differences. As an observation framework, it is still interesting, and the author plans to write about it separately later.
The point of the entire review is not “I made 30% with a dashboard, so you can too.” The more useful message is that a tool can turn feelings into a framework, and a framework into discipline. Making money in one round does not mean a secret method has been found. It can simply mean that the judgment happened to be right this time.