Crypto markets are speculative, emotionally driven, and structurally different from equities or futures. For a systematic trader, that combination is attractive. Where humans trade on emotion and narrative, quantitative methods have room to find edges that more efficient markets have eroded. Brian Blandin from Market Science has been systematically researching those edges since 2017, and his approach is grounded in the same rigorous statistical framework you’d apply to any market.
Brian’s background is mechanical engineering and data science. He got into trading around 2015 and gravitated toward crypto in 2017 specifically because it lacked fundamental anchors. An asset driven mainly by speculation and sentiment creates the kind of mispricings that systematic traders can exploit. His work covers momentum, time-of-day effects, machine learning approaches, volatility management, and the practical challenges of backtesting in a market where flash crashes and liquidation cascades are a normal feature of the data.
This episode is one of the most practical crypto trading discussions the podcast has covered.
Watch the full episode below, then read on for the complete breakdown.
What you need to understand before trading crypto
Before building any strategy, Brian identifies two structural features of crypto that differ significantly from traditional markets.
The first is fragmentation. There is no single exchange. Dozens to hundreds of venues operate simultaneously, each with its own price discovery. There’s no NBBO equivalent in crypto: no single national best bid or offer. Bitcoin trades at slightly different prices on Coinbase, Binance, and smaller offshore venues at the same moment. For data gathering, Brian’s recommendation is to use the exchange you plan to trade on, or to aggregate from two or three of the largest, most liquid venues.
The second is counterparty risk. If a crypto exchange gets hacked or collapses, funds are generally unrecoverable. This isn’t theoretical: there have been multiple high-profile incidents where exchange users lost everything. Brian’s practical response is to keep only a fraction of total capital on any given exchange at any time. With leverage available at 20x or 50x on many platforms, you can take full-size positions while keeping only a portion of your capital exposed to exchange risk.
On data history: Brian uses 2017 as his starting point for most research. That’s when perpetual swaps exchanges became the dominant liquidity venue and when liquidation mechanics began shaping price behavior. Before that, the market structure was different enough that older data has limited relevance. That leaves roughly four years of sample data for most research, which is short by traditional backtesting standards. Brian’s argument: Bitcoin has compressed four to five complete bull and bear cycles into that period. The pace of events is faster than in equities, which means shorter data history still contains meaningful information.
The most reliable edges in crypto
Brian describes momentum as the foundational edge in crypto. The mechanism is straightforward: rising prices attract attention, which attracts more buyers, which causes further price rises. This phenomenon can persist for extended periods before reversing. Selling begets more selling through the same dynamic, amplified by leveraged participants liquidating positions as losses mount.
Beyond momentum, Brian’s firm Market Science focuses on what he calls price anomalies: things that happen on a routine and repeatable basis. These include:
- Time of day effects. The UTC midnight open is consistently the highest-volume hour of the day, driven by automated strategies and daily-timeframe rebalancing orders. During the American and European trading sessions, volatility is elevated. After those sessions, it drops.
- Arbitrage. Because dozens of prices exist for the same asset simultaneously, cross-exchange price differences create arbitrage opportunities. The edge has compressed as sophisticated firms enter the space, but it still exists.
- Candle-based breakout patterns. Brian described a three-candle triangle breakout: when the daily high is lower than the three-day high and the daily low is higher than the three-day low, set a breakout order at the three-day high. He’s used this strategy successfully since he started testing.
On the question of whether older techniques still work in crypto: Brian’s view is yes, at least partially. Simple moving average crossovers and Donchian channel breakouts from trading books written decades ago still produce results when applied to crypto. The market hasn’t yet reached the efficiency level of equities, where those techniques have been largely arbitraged away.
Machine learning approaches to crypto edge-finding
The majority of Brian’s current work uses machine learning rather than rule-coded strategies. The shift reflects a practical observation: specifying patterns through human logic is limited by human imagination. ML approaches let the data surface relationships that wouldn’t occur to a human analyst.
The inputs Brian feeds into his models include price behavior metrics, sentiment data, liquidation data, and open interest. These variables are selected based on whether they have theoretical reasons to be predictive, not because they happened to correlate well in one backtest. The statistical learning algorithm then identifies which combinations of those inputs have persistent predictive value.
For traders without ML expertise, Brian’s starting point recommendation is Build Alpha software (from Dave Bergstrom, who has appeared on BST). It tests candle structure patterns and other rules-based approaches systematically without requiring custom programming. It’s a practical bridge between intuitive strategy ideas and rigorous statistical testing.
Backtesting challenges: slippage and liquidation cascades
This section is worth reading carefully if you’re planning to test crypto strategies.
Standard backtesting assumes you get filled somewhere close to where prices appear in your data. In crypto, that assumption breaks down in specific ways. Flash crashes are a normal feature of the data, not data errors. A price might drop 30% intraday and recover within hours. If your backtest treats that as a valid fill, your results are wrong.
The mechanism behind these events: when a large number of leveraged participants hold positions in the same direction, and prices move against them, exchange liquidation algorithms automatically close those positions by placing market orders. There’s no counterparty on the other side. Prices cascade lower until a large enough buyer steps in. If your stop loss sits in that range, you don’t get filled at your stop. You get filled wherever liquidity finally returns, potentially much worse.
Brian’s recommendation: be generous with slippage assumptions, especially for breakout strategies where you’re entering on momentum. Don’t assume a few basis points. Assume something that accounts for the real distribution of execution quality across market conditions, including the bad ones.
There’s also a stop-hunting problem specific to crypto. Some exchange insiders allegedly have visibility into order flow and can identify where large clusters of stops sit. This doesn’t mean stops are useless, but it’s another reason to not place stops at obvious technical levels and to be conservative about slippage assumptions.
Volatility as a feature, not a bug
Crypto’s annualized volatility runs around 100%, roughly five to ten times higher than equities. Many traders outside crypto see this as a reason to avoid the market. Brian’s view is the opposite: volatility is what makes trading possible. A market that doesn’t move can’t be traded profitably.
The volatility that creates problems is the liquidation cascade variety: sudden 30-50% drops in 24 hours driven by forced selling. These events are different from normal price volatility. The practical response is to understand that these events are in the data and will occur in the future. Building strategies that assume smooth fills during these events is a guaranteed way to underestimate real-world performance.
The size of crypto volatility also means that position sizing should be conservative relative to account size. Even with 100% annualized vol, a well-sized position doesn’t require large capital commitment. The leverage available on crypto exchanges amplifies returns without requiring large capital on exchange, which also reduces counterparty risk exposure.
Selecting which cryptocurrencies to trade
The question of which coins are worth trading comes down primarily to liquidity. For most of the thousands of available tokens, liquidity is insufficient to trade with any meaningful size without significant market impact. A $20,000 to $50,000 order in a thin market can move the price by several percentage points.
Brian’s estimate: at any given time, 20 to 50 tokens have enough liquidity to trade actively without excessive market impact. Beyond the top coins by market cap, he uses a screener that evaluates predicted direction, predicted volatility or range, and predicted trendiness. This narrows the universe to the coins with the best current trading characteristics.
A more recent addition: a sector or theme overlay. Capital in crypto tends to rotate between themes. NFTs and gaming tokens dominated activity in 2021 when this episode was recorded. DeFi was the dominant theme the prior year. Staying aware of which sector is attracting narrative and capital flow lets you focus research on the coins most likely to show statistical edges in the near term.
Related episodes
- Algorithmic crypto trading with Pavel Kycek
- Systematic crypto trading strategies with Adrian Reid
- Quantitative trading strategies: Ernie Chan on momentum, drawdown and automation
Want to learn more about building systematic strategies for crypto markets? Subscribe to the Better System Trader podcast for weekly interviews with the world’s top systematic traders and quantitative researchers.
