Table of Contents
Pavel Kycek—quant-minded founder behind Roboxio—sits down for a candid check-in on building and scaling fully automated crypto trading. In this interview, Pavel explains how his team grew from a tiny shop to an eight-figure, institution-grade managed-accounts operation, why they keep core models stable, and how they engineered a cloud-first trading engine to survive API quirks and 24/7 markets. He also contrasts institutional patience with retail panic and shares how diversification and measured short exposure helped during the early-2025 dump—context beginners rarely get straight from a practitioner.
Read on to learn the pillars of Pavel Kycek’s approach: treat crypto as a momentum-biased market with mean-reverting shocks, diversify widely across coins and model types, size for left-tail risks, and automate with guardrails (alerts, checkers, staged rollouts) so your strategy survives reality—not just backtests. You’ll see why he favors algorithmic trading for most traders, how to balance “don’t touch it” stability with careful iteration, and what it actually takes to manage other people’s money without losing your edge or your nerve.
Pavel Kycek Playbook & Strategy: How He Actually Trades
Core market model: momentum up, mean-reverting dips
Crypto behaves like an up-biased momentum market where selloffs often snap back. Pavel Kycek builds models around that asymmetry and avoids pretending crypto trends equally well both ways. That starting point keeps the system simple and aligned with how the tape actually moves.
- Trade long with trend/momentum/breakout triggers; prioritize the long side as your default edge in crypto.
- Build mean-reversion rules for dips (toward the long side); require volatility compression or intraday exhaustion before entry.
- Avoid pure “momentum short” on major coins—historically the weakest edge; shorts get whipped more often than they trend.
- If you are short at all, favor tightly-defined mean-reversion shorts with smaller size and faster exits than longs.
Portfolio construction & diversification across coins and models
The engine is a portfolio, not a single hero strategy. Pavel splits capital into many small sleeves across low-correlated models and multiple coins, so any one model or coin can be wrong without wrecking the day.
- Split deployable capital into 10–20 equal sleeves; each sleeve runs a distinct model/coin combo. Aim for the lowest cross-sleeve correlation you can achieve.
- Cap per-coin exposure to 5–10% of equity and per-model exposure to 15–25% to avoid model/asset concentration.
- Maintain a model mix (e.g., momentum-long, breakout-long, mean-reversion-long; optional small sleeve of mean-reversion-short).
- Rebalance sleeve sizes weekly or on >50% volatility regime shifts; retire sleeves that add correlation without improving portfolio MAR.
Long vs. short: asymmetry rules that keep you alive
Longs are the workhorse. Shorts exist, but the left-tail risk in crypto (violent upward gaps) is real, so sizing and selection are stricter on the short side.
- Keep gross short exposure ≤ 20% of equity and net exposure long-biased except during extreme downtrends confirmed by regime filters.
- Don’t short illiquid/micro-cap coins; restrict shorts to high-liquidity majors with robust borrow/perp depth.
- Enforce “catastrophe stops” on shorts (e.g., max adverse gap allowance) to respect overnight +300% squeeze risk.
- Prefer time-based exits for shorts (hours/days) and price-based exits for longs; measure edge separately by side.
Execution engine: automation with guardrails
Pavel’s stack treats reliability as part of the edge. Changes roll out gradually, every account is benchmark-checked from the cloud, and a web of alerts watches the bots so nothing drifts without a ping.
- Run a per-sub-account “shadow bot” that reconciles live orders/positions vs. a cloud benchmark; alert on any divergence > X bps or missing fills.
- Canary releases only: push any code/config change to a tiny test account first, then to a few accounts for a few days, then to production.
- During canary, accelerate decision frequency (e.g., 1-hour cycle) to gather feedback fast before wider rollout.
- Build an “alerts above the automation” layer (latency, API errors, position drift, fill anomalies, P&L shocks); never rely on the trading script alone.
Risk architecture: many small bets, many venues, fail-safes
The program minimizes single-point failure—portfolio, exchange, or operational. That means lots of small positions, multi-exchange deployment, and safekeeping policies outside the trading loop.
- Keep position risk tiny (e.g., 0.25–0.75% of equity per position) to absorb coin-specific blowups; prefer 10–20 simultaneous positions.
- Spread accounts across multiple exchanges; keep idle capital off-exchange (cold storage) and trade only highly liquid pairs.
- Enforce portfolio-level drawdown brakes (e.g., cut gross exposure 50% after −10% DD; flat at −15% until recovery protocol clears).
- Separate signal, execution, and oversight processes so one bug can’t cascade across the book.
Research cadence: stable cores, cautious iteration
Edges don’t need constant tinkering; they need careful validation. Pavel keeps core models stable and treats any change like a mini-release with testing, observation, and only then broad deployment.
- Document hypotheses before coding; define failure/retire criteria (e.g., MAR < prior model’s 50th percentile for 6 months).
- Use walk-forward and cross-coin validation to avoid coin-specific overfit; target robustness across regimes, not best backtest CAR.
- Promote changes only after the canary-stage PnL/latency/fill quality matches the benchmark; otherwise, revert automatically.
- For solo traders, keep a tiny “test account” to trial updates before touching the main account.
Operating principles when managing outside capital
Institutional-style patience and process are the edge behind the edge. The team invests heavily in software engineering so the system trades what was designed—no silent drift, no surprise behavior.
- Budget more resources to engineering and monitoring than to inventing ever-new signals; reliability compounds returns.
- Keep a strict change-management pipeline (spec → build → canary → evaluate → deploy), with rollback one command away.
- Communicate risk first: explain why the program is diversified, long-biased, and conservative on shorts.
- Teach the framework: crypto is a trading asset class, not a passive one—design your program accordingly.
Daily/weekly routines that keep the machine sharp
Most of this job is watching the watchers. The routine is about monitoring, reconciling, and staying diversified across venues while keeping idle capital safe.
- Daily: check alert dashboards first, then reconcile positions vs. the benchmark for a random sample of accounts.
- Weekly: re-estimate correlations between sleeves; prune or resize any sleeve that lifts correlation without improving returns.
- Ongoing: rotate exchange exposure to minimize venue risk; cold-store unused balances; restrict trading to high-liquidity pairs.
- After any code/config change: run the canary protocol and 1-hour cycle evaluation window before promoting to all accounts.
Education & skill building for systematic traders
Pavel doesn’t separate trading from engineering or risk; he teaches them together. The goal is a robust, uncorrelated portfolio that can survive real-world exchange quirks and crypto’s 24/7 grind.
- Train on algorithmic foundations and portfolio thinking, not one-off signals; favor uncorrelated strategy sets over “the one setup.”
- Practice building alerting/monitoring layers as first-class features of any bot you ship.
- Treat liquidity, borrow, and exchange/venue risk as model inputs, not afterthoughts; size positions accordingly.
Size Small, Survive Big: Position Risk That Absorbs Crypto Chaos
Pavel Kycek hammers one idea again and again: tiny positions keep you in the game when crypto goes feral at 3 a.m. He sizes trades so a single coin blow-up is a shrug, not a crater, letting the portfolio breathe through drawdowns without panic resizing.
He treats “sub-1% per idea” as a default mindset, then earns the right to scale only when volatility contracts and correlations drop. Stops are placed where the thesis is wrong, not where the pain threshold sits, and he avoids clustering big bets across highly correlated coins in the same regime. When markets accelerate, Pavel cuts unit size first, not last, because survival math beats ego every time. He also separates entry conviction from position size—great setup or not, sizing follows volatility and liquidity, not excitement. The goal is simple: lots of small, uncorrelated swings that compound, while single names can implode without threatening the ship.
Allocate by Volatility, Not Gut: Scale Exposure as Regimes Shift
Pavel Kycek treats volatility like a speed limit: the faster the market moves, the smaller each position gets. He sizes units off realized volatility (think ATR or rolling standard deviation), so exposure scales automatically instead of with emotions. When volatility spikes across the board, he doesn’t debate—he cuts gross exposure and shortens holding time until the tape cools. In calmer regimes, he allows bigger units and wider stops, but only after the numbers confirm compression, not because it “feels” quiet.
He also uses regime filters to keep the portfolio long-biased when momentum is healthy and dialed back when chop or broad risk-off hits. Per-coin caps and a portfolio-level VAR limit stop any single asset from hijacking the day, even if volatility math says “size up.” If daily realized volatility doubles, he halves unit size as a rule, then revisits only after a multi-day cool-down. The result is boring by design: smoother equity, fewer forced decisions, and allocations that adjust faster than any gut ever could.
Diversify by Coin, Model, and Timeframe to Cut Correlation
Pavel Kycek builds the book like a bundle of small engines that rarely fire in unison. He spreads risk across majors and liquid alts, mixes momentum, breakout, and mean-reversion logic, and staggers holding periods from intraday pops to multi-day swings. If two ideas start behaving like twins, he prunes one or shrinks both—correlation is treated as a cost, not a curiosity. Caps per coin and per model keep any single theme from owning the P&L, even when it’s hot.
Time diversification is just as deliberate for Pavel Kycek: short-cycle systems harvest noise, while longer-cycle systems ride the drift, and their wins and losses offset. He rebalances sleeves on a schedule and after regime shifts, measuring who actually helps the portfolio MAR instead of who looks pretty in isolation. New models must add independence first and returns second; if they don’t reduce the portfolio’s “all-up-or-all-down” days, they don’t cut. Over time, that discipline turns a volatile market into a steadier stream of uncorrelated edges.
Trade the Rules, Not Predictions: Mechanics First, Opinions Last
Pavel Kycek doesn’t try to outguess the next candle—he follows a checklist that doesn’t care about opinions. Entries trigger only when his conditions align (trend filter, volatility state, liquidity gate, risk per trade), and exits follow their own playbook regardless of headlines. If the setup is 90% perfect but one rule fails, he passes; the discipline is the edge, not the narrative.
Mechanics also decide when to stand down: Pavel Kycek cuts size when realized volatility spikes, pauses when correlation jumps, and stops trading symbols that fail liquidity tests. Every trade has a predefined stop where the thesis is invalid, plus time-based exits to avoid decay in choppy regimes. He logs rule breaches like defects and fixes them in process, not in prediction—because durability comes from systems that fire the same way on Monday as they do in a panic on Friday.
Define Short Risk Tighter; Let Longs Work With Trend
Pavel Kycek treats shorts like handling a live wire—possible, but only with thick gloves and a short leash. He sizes short positions smaller than longs, places tighter initial stops, and uses faster time-based exits to avoid getting steamrolled by sudden squeezes. Liquidity and borrow depth are non-negotiable; if a symbol can’t absorb size without slippage, it’s off the short list. Catastrophe protection is explicit: hard stops and gap protocols assume upside explosions, not tidy mean reversion.
By contrast, Pavel Kycek lets longs breathe when the broader regime is up and momentum confirms, because crypto’s drift rewards patience on the right side. He allows slightly wider stops and staged adds on strength, but only after volatility compresses and risk stays within portfolio limits. Net exposure stays long-biased unless a regime filter flips, keeping the book aligned with the market’s structural tilt. The philosophy is simple: be surgical and defensive on shorts, but let well-behaved trends carry your longs without constant interference.
In the end, Pavel Kycek’s edge isn’t a single indicator—it’s the way every piece snaps together under pressure. He sizes small so one coin can’t sink the ship, lets volatility set the speed limit, and keeps a long bias that respects crypto’s structural drift. Diversification is engineered, not hoped for: many coins, multiple model types, staggered timeframes, all capped and rebalanced to keep correlation down. The rulebook, not opinions, fires entries and exits; when one condition fails, he simply doesn’t trade. Shorts are handled like a live wire—smaller, faster, tighter—while longs are allowed to breathe when regime and momentum agree. Around all of it sits an automation and monitoring layer that catches drift, API funk, and rollout mistakes before they become P&L problems.
The takeaway for any trader is wonderfully unglamorous: process is profit. Build a portfolio of small, uncorrelated bets; scale exposure with volatility, not confidence; codify rules so they execute the same on a sleepy Monday and a panic Friday. Protect against venue and operational risks as seriously as you guard against market risk. Use canary releases and alerts as part of the strategy, not an afterthought. Most of all, do what Pavel Kycek actually does—treat trading like engineering, so your strategy survives the messy world it has to trade in.
