To make simulator work, one needs just Python, NumPy and Matplotlib installed

Use scripts located in data/fetch-*.py to download price data from Binance. For example, to download CRV/USDT price data:

cd data
python3 fetch-crvusd.py
mv crvusdt.json crvusdt-1m.json  # Format we used in scripts, change to any naming you wish
gzip crvusdt-1m.json

All price data files are assumed to be stored gzipped.

Algorithm is pretty insensitive to AMM fees, however optimal value for almost all coins seems to be within 0.6%-0.9% with the exception of EUR (in examples).

Let's look at fee calculation in example_crv.

python3 1_simulate_fee.py

This script calculates losses depending on the AMM fee for a 4-band position with A=50 (you can change that).

This gives this graph:

The graph suggests that optimal fee is at the minimum of the graph and is equal to 0.6%.

Amplification (A) depends on asset volatility. Let's put the optimal fee we have found into the script 2_simulate_A.py and run it.

We obtain two graphs - losses and discount (sum of losses and the value of liquidity if price slowly goes down) for N = 4 bands:

Value of A for the minimum of orange graph is the A we need to set for the market: A = 50.

Let's take the value of the blue graph at A = 50: it is 0.11 (when rounded up). It is the loss we MAY have in the worst case with N=4 which the position will lose if volatility is really bad.

This means, we set liquidation_discount = 0.11.

It's not great to be immediately liquidated once we get there, so let's add 3% margin to it. This gives: loan_discount = 0.14.

Minimum of the orange graph is 0.135..0.14. Let's add those 3% to it - this is 0.17. It means that: LTV_max = 1 - 0.17 = 0.83 - maximum achievable LTV for CRV.

Therefore, maximum possible leverage for CRV is 1 / (1 - 0.83) = 5.9.

User can adjust position losses by adjusting numbers of bands. Let's calculate losses in SL using recent prices for CRV depending on N using script 4_simulate_avg_loss_brief.py:

Graph calculates loss user would experience when borrowing maximum possible amount and waiting for 1 day (on average). For example, on this graph, at N=50, loss is 0.075% per day.