In trading, slippage is one of the most underestimated execution risks. Execution is where most strategies go to die.
In research, execution is often a parameter. In production, it’s the environment. The same signal can look stable in simulation and become untradeable once you account for spreads widening, fills deteriorating, funding turning against you, and liquidity evaporating precisely when you need it.
This post is not about “getting better entries.” It’s about treating execution as a first-class constraint in crypto spot and futures markets.
1) Slippage is not noise – it is selection
You don’t get filled randomly. You tend to get filled:
- when the market is moving against you (adverse selection)
- when liquidity is thin
- when your urgency is high
That means realized slippage is often state-dependent. It gets worse in exactly the regimes that matter most.
Practical implication: A model that ignores adverse selection will systematically overstate edge.
2) Fees are predictable. Everything else isn’t.
Fees are the cleanest part of the cost stack – and still routinely underestimated.
Costs include:
- taker/maker fees (and tier drift)
- spread
- market impact
- queue position (for maker strategies)
- partial fills
- cancels / replace behavior
Even if you are “maker,” you often pay with adverse selection instead of fees.
Rule of thumb: if a strategy’s edge is comparable to fees, it’s not robust.
3) Funding is a position you didn’t mean to take
Perpetuals funding isn’t just a cost – it’s a regime signal and a risk amplifier.
Funding can:
- flip sign abruptly
- spike during stress
- correlate with crowded positioning
If your system is structurally long when funding is structurally positive, you may be paying for exposure you already wanted. If it’s the opposite, funding can become a silent bleed.
Practical implication: funding must be modeled, monitored, and limited – not hand-waved.
4) Spread and liquidity are dynamic
Spreads are not constants. Depth is not guaranteed. And both can change faster than your system can react.
Bad outcomes happen when:
- you size based on “typical” depth
- liquidity disappears on multiple venues simultaneously
- your orders become the market
Practical implication: sizing and execution logic must degrade gracefully when liquidity thins.
5) Latency is a risk factor, not an optimization hobby
You don’t need to be the fastest to trade systematically – but you do need to know when latency turns your orders into donations.
Latency matters because:
- quotes decay quickly in fast markets
- your fills become more adverse as reaction time increases
- cancels arrive too late
Practical implication: measure end-to-end latency and treat it like a monitored risk metric.
6) What we measure in production
In live trading, we care less about what we intended and more about what happened.
Typical execution KPIs:
- realized slippage vs modelled slippage
- fill ratio (requested vs executed)
- maker/taker mix
- average time-to-fill / cancel effectiveness
- adverse selection proxy (post-fill price move)
- funding paid/received and its variance
If these drift, the strategy may still be “right” and still be losing.
Takeaway
Execution is not a footnote.
If you want research to survive production in crypto spot and futures, you have to design with execution realism:
- conservative assumptions
- explicit monitoring
- hard limits
- graceful degradation
A strategy that cannot survive bad execution is not a strategy.
Leave a Reply