Pick a strategy

A strategy decides how Stratum lends your idle balance into Bitfinex margin funding. This guide covers how strategies are structured, how to choose an algorithm for your experience level, and how to test it in paper mode before going live.

1. One strategy per currency

In Stratum a strategy is a single container for one currency. Its capital is split into allocation buckets, and each bucket runs exactly one algorithm on its slice. A shared orchestrator owns the allocation, the global rate floors and period caps, splitting, and reconciliation against your live offers — so each algorithm only has to decide what offers it wants to rest for its own slice.

That same algorithm code runs in both live ticks and backtests, which is why a backtest reflects how the strategy would actually behave. Backtests are modeled against historical Bitfinex funding data and are optimistic — they assume your offers fill at full size and ignore order-book depth and market impact, so treat any modeled figure as illustrative, not a guarantee.

2. Choose an algorithm by experience level

Every algorithm carries an experience tag — Beginner, Intermediate, Advanced, or Expert — and a profile showing its return potential and how quickly capital frees up. The full marketplace lives on the strategies page. Pick by how hands-on you want to be:

Beginner — lazy / rate-tracking

  • Simple FRR. Lends your whole slice at FRR (Bitfinex's Flash Return Rate, the platform's average funding rate). It matches the crowd and fills fast, but never beats the average — when the rate spikes you are already lent at the old rate. No parameters to set.
  • FRR + Offset. Lends at FRR plus a fixed daily premium you choose, re-pricing as the book moves. A touch more than the average, at the cost of slightly slower fills; set the premium too high and offers stop getting taken. One parameter: the offset (default 0).

Intermediate — rate spread and loan-length ladders

  • Rate Spread. Places a row of tiered offers evenly spaced from below FRR up to above it, splitting the slice evenly. Cheap tiers fill fast; pricey tiers wait to catch spikes. Parameters: number of offers (default 5), cheapest tier (default 0.5× FRR), priciest tier (default 2.0× FRR).
  • Maturity Ladder. Splits the slice evenly across several loan lengths (default 2 / 7 / 14 / 30 days) so something is always maturing and returning to reinvest, like a CD ladder. It splits blindly by length and does not chase whichever length pays best. Parameter: the list of loan lengths.

Advanced — adaptive, curve-aware, and event-driven

  • Adaptive Spread. A Rate-Spread-style ladder whose width grows when the rate is choppy and tightens when it is calm. It only ever lends — it is not a two-sided market-maker — and in calm markets collapses back toward plain FRR. Parameters: number of offers (default 5), calm-condition spread width (default ±10%), and reactiveness (default 2.0).
  • Yield Curve. Ranks the live funding curve's loan lengths by what each is paying and lends across the top few, splitting evenly. It chases the best-paying length even if that rate is only so-so, and can lock you in long at a mediocre rate. With no live curve it falls back to a single FRR offer. Parameter: how many top lengths to spread across (default 3).
  • Rate Ladder. Picks the loan length from the rate level: short and liquid when the rate is low, long and locked when it is high. Capital always rests. By default the triggers adjust to the rate's recent average; you can also set fixed daily-rate numbers. Parameters: trigger style (relative or absolute), the high and low triggers, and short / mid / long loan lengths.
  • Carry Optimizer. Rate Ladder's discipline plus a curve-aware long pick: when the rate climbs above its recent average it locks the best-paying long loan length on offer, capturing the term premium, instead of a fixed length. Parameters: the lock-in and stay-short triggers, and short / mid / minimum-lock lengths.
  • Spike Catcher. Holds the slice idle, lending nothing, until FRR jumps above its recent average by the spike threshold, then deploys the whole slice at a long loan length. Reserved cash earns nothing between spikes, and a brief spike may not give time to fill. Parameters: jump needed to deploy (default +30%), and loan length when it fires (default 120 days).

Expert — self-tuning carry

  • Adaptive Carry — the recommended default. Like Carry Optimizer, but it judges the rate against this market's own recent range (a trailing window of roughly 60 days) instead of a fixed trigger, so it self-tunes per market: it locks readily in steady markets and only on real spikes in spiky ones. When it locks it uses a reinvestment-aware "best carry tenor" so it does not over-reward a length that barely beats staying liquid. It also applies an adaptive rate floor — by default it refuses to lend below the 96th percentile of the recent-rate window, resting the offer until the rate rises to meet it. Until it has enough rate history, it falls back to the simpler average-based rule. The floor leaves a little cash idle in calm markets, which is the price of waiting for better rates.
  • Adaptive Ladder. Rather than committing the whole slice to one length, it splits between a long portion and a short, liquid portion, with weights that slide with the rate's position in its recent range — mostly long near the top, mostly short near the bottom. It is never all-or-nothing, so one mistimed tick cannot strand the whole slice. Its edge over Adaptive Carry is small in backtests; choose it for cross-market robustness, not a higher ceiling.

Return and liquidity are potential, not promises. Higher-return algorithms generally lock capital up longer and depend on rate spikes that may not arrive; lazy algorithms give up upside for near-certain fills. If you are unsure, start with the recommended default and adjust later.

3. Split capital into buckets

You are not limited to one algorithm. A strategy can hold several buckets, each running its own algorithm on its own percentage of the strategy capital — for example a Simple FRR bucket for steady fills alongside a Spike Catcher bucket that waits for spikes. The orchestrator allocates each bucket its slice and applies your global rate floor and maximum loan length across all of them.

Some algorithms appear in the marketplace marked Coming soon (for example Volume-Threshold, Sentiment, and Adaptive Spread + Hedging). These are not yet available — they depend on data feeds Stratum does not yet ingest, and a bucket cannot be saved with one. Only the algorithms marked live can be selected today.

4. Test in paper mode

We recommend running every new strategy in paper mode first. Paper mode simulates against live rates without placing real offers, so you can watch how an algorithm behaves in the current market — how often it fills, how long it locks, how much it leaves idle — before committing capital. To connect an account first, see add your Bitfinex API key.

You can also use the calculator to model outcomes from real funding data. As with backtests, those figures are modeled and optimistic — they are a planning aid, not a forecast.

5. Switch to live

When you are comfortable with how a strategy behaves in paper mode, switch it to live. Stratum then places real funding offers through your scoped Bitfinex API key — a key that grants funding and read access only and is rejected if it can withdraw or trade. For how Stratum stores your key and protects your account, see security.

Past performance does not predict future results. You are lending to margin traders and bear the credit risk of Bitfinex's collateral system; Stratum does not guarantee returns. Start small, watch how your chosen strategy behaves live, and scale up only once you trust it.