Home About Approach Research Partners Contact

Our Approach

How We Think About Risk

Every decision at Kelly Lab begins with a single question: what is the maximum we can lose? From that foundation, we build position sizing, research frameworks, and execution systems that protect capital first — and pursue returns second.

The Kelly Criterion

Developed by John L. Kelly Jr. at Bell Labs in 1956, the Kelly Criterion is a mathematical formula for determining the optimal fraction of capital to deploy on any given trade — one that maximises long-run growth while avoiding the ruin that comes from over-betting.

Most traders ignore position sizing entirely, treating it as an afterthought. Kelly Lab treats it as the primary act of risk management. The formula is simple. The discipline to apply it consistently is not.

Why overbetting destroys capital permanently

If you bet 2× the Kelly fraction, your expected long-run growth rate drops to zero. If you bet more, you experience certain ruin over a long enough time horizon — regardless of your edge. Volatility of returns, not just magnitude, determines survival.

Fractional Kelly as a practical safeguard

In practice, Kelly Lab uses a fractional Kelly approach — deploying a conservative fraction of the full formula output. This accounts for model uncertainty, estimation error in probabilities, and the non-stationarity of digital asset markets.

f* = (bp − q) / b
f* Fraction of capital to allocate to the trade
b Net odds received on the trade (profit per unit risked)
p Estimated probability of a winning outcome
q Estimated probability of a losing outcome (1 − p)
POSITION SIZING OUTCOMES
Underbetting (½ Kelly) Safe, suboptimal
Full Kelly Optimal growth
2× Kelly Growth = zero
3× Kelly+ Certain ruin

Risk before return. Always.

The digital asset market rewards those who survive long enough to compound. Most participants focus entirely on upside — Kelly Lab begins every analysis by asking how much of the downside is knowable, containable, and acceptable before a single trade is placed.

Conventional approach

Size positions based on conviction strength or gut feel

Set stop-losses reactively after entry

Chase momentum and react to price action

Treat drawdowns as unavoidable accidents

Measure success purely by return percentage

The Kelly Lab approach

Size positions mathematically using the Kelly Criterion and estimated edge

Define maximum drawdown tolerance before entry — it governs the position

Anchor decisions in onchain data and market cycle context, not price alone

Treat drawdown avoidance as a primary objective, not a secondary constraint

Evaluate performance by risk-adjusted outcomes and capital preservation over time

Automation removes the human liability

The greatest edge in trading is not a smarter model — it is the consistent, emotionless execution of a sound one. Human traders are subject to a predictable and well-documented catalogue of cognitive biases that erode edge at the moment it matters most.

Kelly Lab automates its execution layer not for speed, but for discipline. The algorithm does not feel fear at a drawdown, nor euphoria at a run-up. It executes the pre-defined logic of the research, every time.

  • Loss Aversion
    Holding losers too long Humans weight losses ~2× more heavily than equivalent gains, causing systematic reluctance to cut losing positions at pre-defined levels.
  • FOMO
    Chasing breakouts past optimal entry Fear of missing out drives late entries — well past the risk/reward levels the model identified as acceptable.
  • Recency Bias
    Overweighting recent price action After a winning streak, traders increase size beyond Kelly limits. After losses, they undersize, missing the recovery.
  • Overconfidence
    Inflating edge estimates post-win Systematic overestimation of one's own probability of success, leading to overbetting and Kelly violations.
How Automated Execution Works
Research Signal Received
The research engine identifies a potential trade thesis based on onchain data and market cycle context.
Input
Edge & Probability Calculated
The model estimates the probability of success and the expected risk/reward ratio for the thesis.
Analysis
Kelly Position Size Determined
The fractional Kelly formula computes the mathematically optimal capital allocation. Size is locked — no override is permitted.
Sizing
Risk & Compliance Checks Pass
Before any order is placed, automated checks verify the trade stays within drawdown limits and meets internal compliance requirements.
Safeguard
Order Executed Without Emotion
The algorithm places the order at the pre-defined size. No hesitation. No second-guessing. No emotional override.
Execution
Result Feeds Back Into Model
Trade outcomes are logged and used to refine probability estimates — making the next decision more precise than the last.
Learning

From signal to sized position

Our research loop runs continuously. Onchain data feeds the model, the model informs conviction, conviction governs position size, and execution results feed back into model calibration.

01
Onchain Data Ingestion
Monitor wallet flows, exchange reserves, miner behaviour, and network activity across major chains in real time.
02
Cycle Regime Assessment
Classify the macro regime — accumulation, distribution, expansion or contraction — using onchain and structural market signals.
03
Edge & Probability Estimation
Estimate probability-weighted expected value for each thesis. Incorporate regulatory context as a risk variable.
04
Kelly Sizing & Compliance Check
Apply the fractional Kelly formula. Run pre-trade compliance assertions. Lock position size before execution is triggered.
05
Review & Model Calibration
Post-trade analysis refines probability estimates and model parameters. The loop improves with every cycle.

Discipline over discretion

01

Named after a formula, not a founder

Kelly Lab takes its name from the Kelly Criterion — a mathematical framework, not a person. Our identity is the methodology. The discipline is the brand.

02

Onchain-native research

We do not use price charts as our primary research input. Blockchain data tells a deeper story — one that price typically confirms weeks later. We read the chain first.

03

Regulatory awareness as alpha

In a market where regulatory developments routinely move prices 20–40%, understanding the regulatory landscape is not just compliance hygiene — it is a genuine source of edge.

04

Minimum drawdown as the primary objective

We do not target a return percentage. We target a drawdown ceiling. Everything above that floor — position sizing, entry logic, execution timing — is calibrated to protect it.

05

Proprietary capital only

Kelly Lab trades its own capital. We are not managing your money, and we are not incentivised by AUM growth. Our interests are aligned with outcomes, not assets under management.

06

Built to compound, not to impress

Flashy returns attract attention. Consistent, low-drawdown compounding builds lasting capital. We optimise for the latter, which is mathematically the harder and more valuable goal.

Explore our research or start a conversation

See how our onchain research framework translates into actionable market intelligence, or reach out if our approach resonates with how you think about digital assets.