Betfair Exchange API Bot Coding: Execution Under Minimum Liquidity
Minimum liquidity on the Betfair Exchange is where 90% of bot coders get slaughtered. Not because the API is broken, but because they treat thin markets like liquid ones and wonder why their fill rate collapses to zero while the spread eats them alive. If your bot cannot survive a market with £47 matched on the back side, it does not deserve to touch a market with £47,000.
The Real Problem: Order Book Depth Is a Lie
The Betfair Exchange API returns a price ladder that looks functional. Three back prices, three lay prices, decent gaps. But look at the size behind those prices. In minimum liquidity markets, you will routinely see £2 available at the best back price. Your bot fires a £50 order, gets £2 filled, and the remaining £48 sits there like a dead body, exposed to adverse selection. The API does not warn you. The order book does not scream. It just lets you hang.
This is not a coding bug. This is a market microstructure reality that most automated betting systems ignore completely. The Betfair Exchange API bot coding approach must account for fragmentation, phantom liquidity, and the brutal truth that displayed size is often pulled the millisecond your order hits the queue.
Dynamic Stake Sizing Based on Visible Depth
Stop hardcoding your stake. That is amateur hour. Your bot needs to query the current market depth through the listMarketBook endpoint and calculate a stake that is a percentage of the actual money available at your target price. A 15% cap on visible liquidity is aggressive but workable. A 5% cap is conservative but keeps you invisible to market makers who hunt oversized orders in thin books.
Here is the logic that actually works: pull the available size at your desired price, multiply by a liquidity factor between 0.05 and 0.20, round down to the nearest £2 increment, and fire. If the calculated stake drops below the minimum bet threshold, your bot should skip the trade entirely. No exceptions.
Execution Algorithms for Thin Markets
Iceberg orders are your friend. The Betfair API supports partial matching through its standard order types, but you need to build the iceberg logic yourself. Split your intended position into 4 to 7 child orders. Randomize the timing between them by 300 to 900 milliseconds. Randomize the size of each child order by ±18% to avoid pattern detection. This is not paranoia. This is survival.
The minimum liquidity execution problem also demands that you monitor the fill rate in real time. If your child orders are only getting 40% filled within 2.5 seconds, your bot must automatically widen its acceptable price range or abort the remaining children. Sitting in the queue like a statue is how you get picked off.
Latency Arbitrage in Illiquid Markets
Here is the counterintuitive truth: minimum liquidity markets are where latency arbitrage still works, but only if your bot understands the difference between real liquidity and bait liquidity. Market makers in thin books will post £5 at a price that looks attractive, wait for your bot to take it, then immediately move the market against your position. Your bot needs a confirmation delay of at least 400 milliseconds before committing to any price that appears more than 3 ticks away from the mid.
Cross-reference the traded volume from the last 60 seconds against the current available size. If the ratio exceeds 8:1, the liquidity is real. If it is below 3:1, you are looking at ghosts.
Error Handling That Does Not Kill Your Bot
The API will throw errors in thin markets. ERROR_INSUFFICIENT_FUNDS when you have plenty. ERROR_MARKET_NOT_AVAILABLE when the market is clearly active. These are not bugs. These are the exchange protecting itself from erratic order flow. Your bot must categorize errors into retryable and fatal. Retryable errors get a 1.2-second exponential backoff. Fatal errors trigger a 30-second cooldown on that market. Never let a single error crash the entire execution loop.
Log every rejection with the timestamp, the market ID, the requested price, and the available size at that moment. After 500 trades, run a correlation analysis. You will find that 70% of your rejections cluster around specific price points where the exchange's matching engine is throttling activity. Blacklist those price points for 15 minutes.
The Streaming API Is Not Optional
Polling listMarketBook every 200 milliseconds will get your API key rate-limited in under 4 minutes. You need the Exchange Stream API. Subscribe to the market subscription channel, filter for price changes at the top 3 levels, and process updates in a separate thread from your order execution logic. A single thread doing both will choke when the market moves fast, which in minimum liquidity conditions happens every 8 to 12 seconds.
Buffer your stream updates in a ring buffer of 256 entries. Your execution thread reads from the buffer at its own pace. If the buffer overflows, your bot is too slow to trade that market. Shut it down and move to the next one. There are over 80,000 markets on a typical Saturday. Do not marry a dead one.
Final Technical Reality
Minimum liquidity execution on the Betfair Exchange is not about better code. It is about better market selection. Your bot should scan 200 markets, identify the 12 with the cleanest order book dynamics, and execute only there. The rest are noise. The API gives you the data. Your job is to not be the sucker who provides the exit liquidity for everyone else.