DeFI Decode: a strategy suitable for retail users with higher-return less-drawdown compared with just-hold BTC (Part I)

5 min readJan 29, 2024

The purpose of this article is to investigate the problem of whether we can design a strategy suitable for retail users, with higher-return, less drawdown and infrequent execution, compared with just-hold BTC. And if have, how robustness is such a strategy. (The Github code is here)

This will be a series of 4 parts:

Part I: Momentum based strategy and choice of optimal lags (this article)
Part II: Evaluation of the statistical significance of momentum trend
Part III: Risk reduction through signal strength and stop loss control
Part IV: Adding statistical arbitrage (against ETH) to build market neutral strategy.

This article is the Part I. In our thought, the retail suitable strategy should meet criteria of:

Infrequent trade and definitely no-intraday trading. This means we will just use end-of-day data
Less drawdown compared with buy-and-hold, due to that deep drawdown is a key reason why retails users cannot hold.
Higher return.
Robustness against execution. Even users do the trade a few days after the signal, the strategy should still work.
No short of market and no leverage. Such is dangerous for retail users in crypto market.
Other maintenance: data should be free and no need any coding skills for producing the signal.

Is there such a strategy?

yes, through test, we find out that momentum based strategy with 30 day look back period (mmt_30) meets all above criteria. In short:

when past 30-day return is positive, buy BTC.
when above is negative, sell it out and keep cash.

Dose it really that simple? well, only data can tell.

In data of past 5 years, mmt_30 produces 25ppt higher return with 20ppt less drawdown. and longer data shows similar results!

Let’s get into details.

It may be the race is not always to the swift, or the battle to the strong — but that’s the way to bet.
— Damon Runyon (I learn from book “What works on wall street”)

Summary of the Return/Risk results:

We use past 5 years data for this article (and longer data from 2015 gives the same results).

Mmt_30 achieves 25ppt higher compared with just_hold

Chart below shows mmt_30 beats just_hold strategy yearly in the past 5 years. Note that just_hold is hold for whole year, while mmt_30 holds and park cash depends on the price momentum. So case like year 2023 when price mostly are recovering, the 2 strategy produce similar outcome. But mmt_30 produce much better results when price swings like 2021 and 2020.

In terms of numbers, mmt_30 (yearly average at 133%) is 25ppt better compared with just_hold (108%).

Mmt_30 achieves 20ppt LESS in drawdown compared with just_hold

The yearly max drawdown are in chart below.

In terms of numbers, just_hold average drawdown is -48%, while mmt_30 is only at -28%. The significant drawdown (most above 50%) is main reasons retail users having challenges to hold. Here, i believe the improvement in this factor is critical for overall better yield for retail users.

Other risk-return metrics includes Sharpe and Sortino ratio. they are produced in the table below as well for reference.

Ths justification & robustness of 30 day lookback

Lookback period has 2 impact:

Will the performance similar or change significantly around 30-day period,
Retail users may only loosely follow the 30day lookback signal.

The chart below shows the mmt_return (exact look-back period), roll_avg_mmt_return (5 days before/after + current day = 11-day window), with the just_hold_return as benchmark.

you can see:

Highest return happens between 25 days to 40 days.
When lookback < 10 days, the return drops below just_hold, showing frequent trading leads to lower return.
Also it is interesting to see that the local peaks happen at ~30days, ~60days and 120days.