Skills

Backtest a Trading Strategy Honestly

5 weeks · 0 milestones

Write rules first, backtest with costs and no look-ahead, then test on a held-out period and defend the results. Historical data only, not investment advice.

Part of: Business & Management · Proves for: Career

Milestone map

3 milestones

Write the rules first

1 week

Write a simple, rules-based strategy down before testing it — entry, exit, position size and rebalance timing. Rules written first stop you from fitting the past.

Proof required

Submit the dated strategy rules and the data you will test on (source, assets, dates).

What gets checked

Every rule is exact, not 'buy when it looks cheap'
Dated before testing
Asset list avoids survivorship bias or says so

Common mistakes

Vague rules
Rules edited after seeing results
Testing only today's winners

Powstik Guide

Exact, dated strategy rules.

Steps

Write entry, exit and sizing rules.
Set the rebalance schedule.
Pick the asset universe.
Note survivorship risk.
Date the rules.

Template

Entry:
Exit:
Size:
Rebalance:
Universe:
Date written:

What gets sent back

Subjective rules.
Undated rules.
Hand-picked winners.

Resources

Foundationstart here

What a verifier looks for

Checked by the verifiers you invite to your panel (up to 3). Pick people who can judge the work, not friends.
Check each rule could be coded exactly
Ask how the asset list was chosen

You'll sign in first, then come straight back here.

Backtest with real costs

2–3 weeks

Code the backtest with no look-ahead, include trading costs, and use walk-forward testing. Report return, volatility, Sharpe ratio and maximum drawdown.

Proof required

Submit the backtest code and a results table with and without costs, including Sharpe and max drawdown.

What gets checked

Signals use only past data
Costs included
Drawdown reported, not just return

Common mistakes

Using same-day close to trade at that close
Zero costs
Only showing total return

Powstik Guide

An honest backtest with costs.

Steps

Lag signals by one period.
Add a cost per trade.
Walk forward through time.
Compute Sharpe and max drawdown.
Compare to buy-and-hold.

Template

Metric | no costs | with costs | buy-and-hold
Return
Volatility
Sharpe
Max drawdown

What gets sent back

Look-ahead in signals.
No costs.
No drawdown.

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

Checked by the verifiers you invite to your panel (up to 3). Pick people who can judge the work, not friends.
Check for look-ahead in the signal
Compare with and without costs

You'll sign in first, then come straight back here.

Stress it and defend it

1–2 weeks

Test the strategy on a period and assets you left out, and change one parameter to see if results hold. Defend it live to a reviewer.

Proof required

Submit results on the held-out period, a parameter sensitivity table, and the review log.

What gets checked

Held-out period shown even if worse
Results don't hinge on one lucky parameter
Reviewer questions answered

Common mistakes

Hiding the held-out result
Tuning until it works
No sensitivity check

Powstik Guide

A strategy tested where it wasn't tuned.

Steps

Run on the held-out period.
Vary one parameter ±50%.
Tabulate results.
Present to a reviewer.
Log the challenges.

Template

Review session log
Date:            Duration:
Reviewer name:
Reviewer role and experience:

Challenge 1 (strategy robustness):
  What they asked:
  My answer:
Challenge 2:
  What they asked:
  My answer:

Change I made after the session:

What gets sent back

No held-out test.
No sensitivity table.
No review.

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

Checked by the verifiers you invite to your panel (up to 3). Pick people who can judge the work, not friends.
Pick a parameter to change
Ask why the edge should last

You'll sign in first, then come straight back here.

Part of

Quant Modeling →

Related outcomes

Milestone map

3 milestones

Write the rules first

1 week

Write a simple, rules-based strategy down before testing it — entry, exit, position size and rebalance timing. Rules written first stop you from fitting the past.

Proof required

Submit the dated strategy rules and the data you will test on (source, assets, dates).

What gets checked

Every rule is exact, not 'buy when it looks cheap'
Dated before testing
Asset list avoids survivorship bias or says so

Common mistakes

Vague rules
Rules edited after seeing results
Testing only today's winners

Powstik Guide

Exact, dated strategy rules.

Steps

Write entry, exit and sizing rules.
Set the rebalance schedule.
Pick the asset universe.
Note survivorship risk.
Date the rules.

Template

Entry:
Exit:
Size:
Rebalance:
Universe:
Date written:

What gets sent back

Subjective rules.
Undated rules.
Hand-picked winners.

Resources

Foundationstart here

What a verifier looks for

Checked by the verifiers you invite to your panel (up to 3). Pick people who can judge the work, not friends.
Check each rule could be coded exactly
Ask how the asset list was chosen

You'll sign in first, then come straight back here.

Backtest with real costs

2–3 weeks

Code the backtest with no look-ahead, include trading costs, and use walk-forward testing. Report return, volatility, Sharpe ratio and maximum drawdown.

Proof required

Submit the backtest code and a results table with and without costs, including Sharpe and max drawdown.

What gets checked

Signals use only past data
Costs included
Drawdown reported, not just return

Common mistakes

Using same-day close to trade at that close
Zero costs
Only showing total return

Powstik Guide

An honest backtest with costs.

Steps

Lag signals by one period.
Add a cost per trade.
Walk forward through time.
Compute Sharpe and max drawdown.
Compare to buy-and-hold.

Template

Metric | no costs | with costs | buy-and-hold
Return
Volatility
Sharpe
Max drawdown

What gets sent back

Look-ahead in signals.
No costs.
No drawdown.

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

Checked by the verifiers you invite to your panel (up to 3). Pick people who can judge the work, not friends.
Check for look-ahead in the signal
Compare with and without costs

You'll sign in first, then come straight back here.

Stress it and defend it

1–2 weeks

Test the strategy on a period and assets you left out, and change one parameter to see if results hold. Defend it live to a reviewer.

Proof required

Submit results on the held-out period, a parameter sensitivity table, and the review log.

What gets checked

Held-out period shown even if worse
Results don't hinge on one lucky parameter
Reviewer questions answered

Common mistakes

Hiding the held-out result
Tuning until it works
No sensitivity check

Powstik Guide

A strategy tested where it wasn't tuned.

Steps

Run on the held-out period.
Vary one parameter ±50%.
Tabulate results.
Present to a reviewer.
Log the challenges.

Template

Review session log
Date:            Duration:
Reviewer name:
Reviewer role and experience:

Challenge 1 (strategy robustness):
  What they asked:
  My answer:
Challenge 2:
  What they asked:
  My answer:

Change I made after the session:

What gets sent back

No held-out test.
No sensitivity table.
No review.

Resources

Foundationstart here

Depthgo deeper

What a verifier looks for

Checked by the verifiers you invite to your panel (up to 3). Pick people who can judge the work, not friends.
Pick a parameter to change
Ask why the edge should last

You'll sign in first, then come straight back here.