Traffic Light System Research Framework

Framework for the Research

• In order to ensure consistency amongst the studies, a framework has to be built which is followed in each research for the given sector.
• This framework will allow us, to compare different studies inside the same sector, to see the differences in their performance; their strength and weaknesses against each other.
• Each sector will have its own framework it follows (or not? thoughts on the last slide).
• All study sectors have to have a fixed timespan in order to compare results (I like to take BTC from 2018, as before that I don't find it relevant for studies due to its volatility).
• The framework has to be built first before the studies, and can’t be edited drastically after a couple studies are done.
• In the next slides I will show my framework for the Trending section.

Framework for the Trending Sector: Data summary

Fixed study period for all studies: 2018.01.01 - 2024.10

Data summary at the end of each study:

1. Hit Rate of the study: Number of winning and losing events
2. Returns: % SUM of all recorded events
3. Average Max Returns: % average of peak % achieved in winning events
4. Average Max Drawdowns: % average of peak % drawdowns in winning events
5. Average Max Returns: % average of peak % achieved in losing events
6. Average Max Drawdowns: % average of peak % drawdowns in losing events
7. Average “win”, and average “loss”
8. SUM of Time spent in the market in days
9. Return (%) Time (Days) ratio = Return SUM divided by Time SUM
10. EV = (average win * winrate) - (average loss * lossrate)

​

Measured data explanation: Why do I measure them?

1. Hit Rate of the study: Number of winning and losing events:

• To see how many times the rules win, and lose

2. Returns: % SUM of all recorded events:

• To see how much % is the study able to capture

3. Average Max Returns: % average of peak % achieved in winning events:

• To see what the max peak of the winning events are in % terms (needs to be compared to the average win to spot big differences = potential flaws of the rules)

4. Average Max Drawdowns: % average of peak % drawdowns in winning events

• To see what the max drawdown of the winning events are in % terms (gives an idea how much the downside is before the market goes in your favour)

5. Average Max Returns: % average of peak % achieved in losing events

• To see what the max upside of the losing events are in % terms (gives an idea how much the upside iis before the market goes against you)

​

​

Measured data explanation: Why do I measure them? Part 2

6. Average Max Drawdowns: % average of peak % drawdowns in losing events

• To see what the max drawdown of the losing events are in % terms (needs to be compared to the average loss to spot big differences = potential flaws of the rules)

7. Average “win”, and average “loss”

• To see the average of winners and losers, needed also to calculate EV

8. SUM of Time spent in the market in days

• To see how much time you spend in the market in total in a given study period = market exposure in time

9. Return (%) Time (Days) ratio = Return SUM divided by Time SUM

• A number based on the returns and time, can easily be compared with other studies to see which captures the most % in the less time = the bigger this number, the better

10. EV = (average win * winrate) - (average loss * lossrate)

• An easy tool to compare studies in their performance based on their wins, losses and their hitrate

​

Framework for any single data point

recorded inside studies of this sector

• Win / Loss
• Return
• Max peak upside
• Max peak downside
• Time in Days spent in the market

In the next 3 slides I show a visual example how one data point will look in any study in this sector:

Example 1:

Study rules: Buy when 1D 12/21 green, exit when 1D 12/21 red

Date: 21.jun.2023

Outcome: Win

Return: +3.05%

Max peak: +12.4%

Max drawdown: -0.11%

Time in Days: 33D

​

Example 2:

Study rules: Buy when there is a daily close above the bollinger channel, sell when there is a daily close inside

Date: 02.dec.2023

Outcome: Win

Return: +10.37%

Max peak: +13.29%

Max drawdown: -0.45%

Time in Days: 8D

​

Example 3:

How will the candle close after 5 consecutive green days?

Date: 29.feb.2024

Outcome: Red

Return: -2.01%

Max peak: +2.02%

Max drawdown: -3.20%

(Time in Days: 1D)

​

Framework of other sectors:

Should the other sectors have other frameworks?

• I thought YES, but thinking deeper, the before mentioned framework could actually work for mean reversion and other sectors too, as it is basically only recording: if the signal was profitable or not (win or loss), how much returns did it had in either case, and how much time did the event last.
• Even studies like: 5 consecutive candles green, how will the next candle close? Can be conducted inside this framework. Instead of the win/loss words we can simply use green/red, but the overall framework stays the same: Green/Red, Return, Max peak, Max drawdown, (the time is in that case 1D always).