The Math Behind SkyDriver 🧮

This document is just a a summary of notes.md

Warning

Math ahead! If you just want to use SkyDriver, you can skip this. But if you’re curious how the predictions work, read on!

Note

Found a problem with the math? Please report it! I’m not a mathematician, so there might be mistakes.

Overview

SkyDriver uses several indicators to predict if an item will be profitable to flip. Each indicator looks at different aspects of the market, and they work together to give us a prediction.

Smoothing Functions

We smooth out the results of the other calculations to make them more manageable. We have a few ways to do this:

No Smoothing

Just use the raw numbers (boring but sometimes useful)

Sigmoid Smoothing

$f (x) = \frac{200}{1 + e ^{- k x}} - 100$

Tanh Smoothing

$f (x) = 100 tanh (k x)$

Saturating Smoothing

$f (x) = \frac{100 x}{1 + k x ^{2}}$

Piecewise Smoothing

For x > 0: $f (x) = \frac{x}{( 1 + ( \frac{x}{100} ) ^{n} ) ^{\frac{1}{n}}}$

For x ≤ 0: $f (x) = - \frac{- x}{( 1 + ( \frac{- x}{100} ) ^{n} ) ^{\frac{1}{n}}}$

Note

$k$ and $n$ control how aggressive the smoothing is

Market Indicators

1. Price Spread (PS)

How different the buy and sell prices are:

$PS = \frac{b u y P r i ce - se llP r i ce}{se llP r i ce} \times 100$

2. Volume Imbalance (VI)

Are more buys or sells going through?

$V I = \frac{b u y V o l u m e - se ll V o l u m e}{b u y V o l u m e + se ll V o l u m e} \times 100$

3. Order Imbalance (OI)

Are there more buy orders than sell orders?

$O I = \frac{b u y O r d ers - se llO r d ers}{b u y O r d ers + se llO r d ers} \times 100$

4. Moving Week Trend (MWT)

What’s been happening over the last week:

$M W T = \frac{b u y M o v in g W ee k - se llM o v in g W ee k}{b u y M o v in g W ee k + se llM o v in g W ee k} \times 100$

5. Top Order Book Pressure (TOBP)

Looking at the top 30 orders:

$TOBP = \frac{\sum _{i = 1}^{30} ( b u y A m o u n t _{i} \times b u y P r i c e _{i} ) - \sum _{i = 1}^{30} ( se ll A m o u n t _{i} \times se llP r i c e _{i} )}{\sum _{i = 1}^{30} ( b u y A m o u n t _{i} \times b u y P r i c e _{i} ) + \sum _{i = 1}^{30} ( se ll A m o u n t _{i} \times se llP r i c e _{i} )} \times 100$

6. Volume Factor (VF)

How much trading is happening:

$V_{t o t a l} = b u y M o v in g W ee k + se llM o v in g W ee k$

$V F = - 100 + (\frac{V _{t o t a l} - V _{l o w}}{V _{hi g h} - V _{l o w}}) \times 200$

7. Profit Margin Factor (PMF)

How much profit you might make:

$PM = se llP r i ce - b u y P r i ce$

$P M_{p erce n t a g e} = \frac{PM}{se llP r i ce}$

$PMF = - 100 + (\frac{P M _{p erce n t a g e} - P M _{l o w}}{P M _{hi g h} - P M _{l o w}}) \times 200$

Putting It All Together

The Final Prediction

We combine all these indicators with different weights:

$P_{p re d} = w_{1} \times PS + w_{2} \times V I + w_{3} \times O I + w_{4} \times M W T + w_{5} \times TOBP + w_{6} \times V F + w_{7} \times PMF$

Where all the weights ( $w_{1}$ through $w_{7}$ ) add up to 1.

Note

The weights are configurable in the Config if you want to change how the predictions are weighted

What The Numbers Mean

Positive number = Probably profitable
Negative number = Probably not profitable
Higher % = Stronger prediction

Confidence Score

We also calculate how confident we are in the prediction:

$C o n f i d e n ce = \frac{Number of indicators agreeing with prediction}{7} \times 100$

Tip

Want to see how this works in code? Check out internal/priceFluctuation

☁️ SkyDriver

Explorer

The Math

The Math Behind SkyDriver 🧮

Overview

Smoothing Functions

No Smoothing

Sigmoid Smoothing

Tanh Smoothing

Saturating Smoothing

Piecewise Smoothing

Market Indicators

1. Price Spread (PS)

2. Volume Imbalance (VI)

3. Order Imbalance (OI)

4. Moving Week Trend (MWT)

5. Top Order Book Pressure (TOBP)

6. Volume Factor (VF)

7. Profit Margin Factor (PMF)

Putting It All Together

The Final Prediction

What The Numbers Mean

Confidence Score

Graph View

Table of Contents

Backlinks