Teaching example
Made-up inputs help explain a decision and its costs. They are not observations from a market.
START HERE / A SHORT WALKTHROUGH
Would this rule
buy or wait?
Follow one decision, then see why a price gain can still lose money. That is the kind of question we investigate at CouchChange.
A fictional asset, no market data or real orders. This simplified rule is for learning; it is not one of our registered strategies.
01 / READ THE INSTRUCTIONS
Buy only when both conditions pass.
An algorithm is a set of instructions a computer follows. For this example, the price must be up at least 2% over the past day and must not be falling over the past hour.
Past-day price change+3.00%The rule requires at least +2%.
Past-hour price change−0.50%The rule requires zero or above.
Why check both? The daily check asks for a sustained rise. The hourly check avoids buying during a recent dip. Those choices may miss opportunities or still lose money; they do not prove the rule is useful.
Read the answer and explanation
Wait. The +3% daily rise passes, but the −0.5% hourly change fails. Both checks must pass. Correctly following this rule tells us nothing yet about whether it makes money.
02 / A DECISION IS NOT A COMPLETED TRADE
Suppose the next check passes.
In our invented next observation, the day is still up 3% and the hour is now up 0.25%. Both conditions pass, so the rule requests a buy.
- Order requestedThe simulated budget is $100, including the buying fee.
- Fill assumedFor this example, we assume the purchase completes. A “fill” is the part of an order that executes.
Paper trading uses simulated orders and pretend money. Real orders can arrive late, fill only partly, or fail. A request to buy is not proof that we bought at the expected price.
03 / FOLLOW THE MONEY
The price rises 1%. We sell. Did we profit?
Our example sells everything after that made-up rise. The $100 budget includes the entry fee. We charge a fee on both the buy and the sell, plus a small allowance for receiving a worse price than expected.
Result after modeled costs-$0.70
A 1% price gain still loses money under these assumptions.
- Proceeds from the sale
- $99.30
- If the full $100 had no costs
- +$1.00
- Price rise needed to break even
- 1.71%
Teaching example, not a forecast, exchange quote or replay of a CouchChange trade. It omits spread, depth/size impact, network fees, delays, failed orders, taxes and hosting. Real results may be worse. Changing these sliders does not change any study.
What did we learn?
With the starting assumptions, a 1% price rise returns $99.30: a loss of $0.70. The costs outweighed the gain. One invented trade demonstrates the arithmetic; it cannot tell us whether a strategy will work.
04 / KNOW WHAT YOU’RE READING
Examples, designs and results are different.
Study design
The actual question, rules and limits of an experiment. A design alone says nothing about its performance.
Recorded results
Dated observations and simulated outcomes. Our recorded evidence remains private while publication rights are checked.
Where does AI fit? AI helps us write software, propose questions and review evidence. Our current paper trades follow fixed rules. An LLM, or large language model, can explain an idea convincingly and still be wrong about it.