Test the Theory | Research through October 15, 2026 | Published October 17, 2026
A historical daily-chart study using data since January 2020. These are dated examples, not current market setups.
Does waiting improve the entry?
A stock breaks out. You wait for it to come back, it holds the old resistance, and then it runs. That is the version of the break and retest we usually see on a chart. But what about the retests that fail, and the stocks that never come back? Does waiting actually help? For this first episode, we tested NVIDIA, Tesla and AMD using daily prices from January twenty twenty through this October. The chart examples are all from the past year.
Three entries at a level identified at the time
Here is the setup. A stock has to close above the highest price from its previous twenty trading sessions. That gives us a level we could have identified at the time. We then compare three approaches: buy at the next open after the breakout; buy the first return to that old high; or wait for price to test the level and finish the day above it, then buy the following open. That last one is the daily confirmation I like to see. A touch during the day is not the same thing as a daily close above. We give the retest ten trading sessions to happen. If it never qualifies, that approach stays in cash.
The same exit for every approach
All three approaches get the same exit date, thirty trading sessions after the original breakout. We only include setups with that full follow-through available. We leave space between setups on each stock so the same run does not get counted over and over. These are price returns, so dividends are left out. We include a small execution cost on each side, and also check twice that cost. There is no stop or profit target in this test. We are isolating the entry rule, not presenting a complete trading system. The first-touch entry is an idealized limit-order model; seeing a daily low touch a level does not guarantee you would have received a real fill.
AMD: a confirmed winner in May 2026
Start with AMD in May this year. It closed above the old high, came back to that area, and finished the retest day above it. The confirmed entry happens the following morning, after we actually know that close. By our shared exit in June, that entry was up about fourteen percent after the assumed costs. Buying the breakout gained about thirteen percent, while the first-touch model gained about twenty-one. Confirmation worked here, but waiting for that close meant paying more than the first-touch entry. The cleanest-looking signal was not the cheapest entry.
NVIDIA: confirmation still lost in December 2025
Now compare that with NVIDIA last December. The breakout appeared, the level was tested, and the stock finished the day back above it. The next regular session opened higher, so the confirmed entry paid more. By the common exit in February, confirmation had lost roughly two and a half percent. Buying the breakout and the first-touch model each lost roughly one and a half percent. The daily close above the level did not prevent a losing trade. We chose these examples using a fixed selection rule within the past year, rather than showing the biggest winner and the worst collapse. Those recent example groups are small.
NVIDIA: the April 2026 move that never retested
NVIDIA in April gives us the other side of waiting. After that breakout, it never returned to our level during the next ten sessions. Both waiting approaches stayed out. Buying the breakout gained about eight percent by the shared exit in May. Avoiding a bad entry can help, but missing a move also belongs in the results. A chart of the trades we took cannot tell us what happened to the opportunities we passed on.
All 94 completed setups since 2020
Across all three companies, we had ninety-four completed setups since twenty twenty. Buying the breakout entered every time. The first-touch model entered sixty-two times. Waiting for daily confirmation entered forty-one times and stayed out on fifty-three. Across all the original setups, with those no-trade cases counted as zero, the average result was a little over eight percent for the breakout, about four and a quarter percent for the first touch, and about four percent for confirmation. That average is per setup, not an annual return or a portfolio result. The breakout also had the higher average on each company separately. Doubling the assumed costs kept that overall ordering unchanged.
Entered trades and missed opportunities
Here is where the story can become misleading. If we show only the confirmed trades that actually entered, their average was a little over nine percent. That sounds stronger, but it leaves out more than half of the original opportunities. Across all opportunities, the median result was zero for both waiting approaches because so many never entered. We should show the trades and the missed trades together, rather than letting the entry filter hide them.
Limits of this three-company pilot
This does not mean buy every breakout. These are three well-known surviving stocks, and we already know they became major names. That selection matters. The stocks are also related, so these are not fully independent experiments. We have not tested a broad market universe, every retest definition, or which approach controls risk best. Even the small twenty twenty-six slice looked different: waiting had a higher pooled average, but there were only nine completed setups. That is too little to call a reliable change. For this particular rule and sample since twenty twenty, waiting for confirmation did not improve the average result across all opportunities. A setup can look cleaner on the chart and still give up enough trades to change the outcome.
More setup research
The full rules, results and charts will be on YieldTerminal Pro. Follow the channel and our X account for the next test. And remember, this is market research only, not a recommendation to buy or sell. Thanks for watching. I will catch you in the next one.
Sources and study boundaries
Daily price history: NVIDIA, Tesla, AMD. The study uses completed daily sessions, a prior 20-session high, a 10-session retest window and a common exit 30 trading sessions after the breakout. Full follow-through is required. Costs assume 5 basis points per side with a doubled-cost sensitivity check. Returns exclude dividends and are event returns, not a portfolio or a trading recommendation.
These three surviving companies were selected examples; correlated events, idealized first-touch fills and the absence of stops limit generalization. Daily confirmation requires the retest day to close above the prior level. A chart touch alone does not establish an executable fill.