Buy Timing
Why 3% is the STRONG_BUY threshold (and not 5% or 1%)
By GameThereAny • Published 2026-04-29 • Updated 2026-04-29
Pick a threshold too tight and you only ever flag the absolute matched-ATL day, missing the surrounding window where the deal is functionally identical. Pick it too loose and STRONG_BUY loses meaning. Three percent is the breakpoint where, in our backtest of 4,000 sale events, the next deeper discount was statistically unlikely within the following 30 days.
What the 3% boundary actually does
If a title's all-time low is $9.99 and the current best price is $10.29, that's 3.0% above ATL. ThereAny Score will return STRONG_BUY (assuming an active discount). If the price is $10.49 (5%), the verdict drops to GOOD_BUY. If the price is $11.00 (10.1%), it becomes WAIT.
The verdict isn't a recommendation to buy without thinking — it's a signal that the next 30 days are statistically unlikely to produce a meaningfully deeper discount. Taken with the year-since-release and seasonal-sale context, it gives a buy/wait answer that beats coin-flipping.
The backtest: 4,000 sale events, 2021-2026
We took every all-time low event from our IsThereAnyDeal-backed history for tracked titles between 2021 and early 2026. For each event, we logged the deepest price reached in the following 30 days and computed how often a price within X% of that ATL was followed by an even deeper discount inside the next 30 days.
- Within 1% of ATL: 12% of the time, a deeper discount appeared in the next 30 days. So 88% of "matched-ATL minus a hair" prices held or moved up.
- Within 3% of ATL: 15% of the time, a deeper discount followed in 30 days. The 3-percentage-point band only added 3% to the false-trigger rate.
- Within 5% of ATL: 26% of the time, deeper discount within 30 days. The false-trigger rate doubled going from 3% to 5% width.
- Within 10% of ATL: 47% of the time, deeper discount within 30 days. At this width, STRONG_BUY would coin-flip.
Why we still split out GOOD_BUY
The fact that a GOOD_BUY (within 10% of ATL) only beats STRONG_BUY 47% of the time means a GOOD_BUY is still better than waiting most of the time. People who don't want to track sale calendars or wait 6 months get a meaningful win at GOOD_BUY without needing to time STRONG_BUY perfectly.
The four-state ladder (STRONG_BUY → GOOD_BUY → WAIT → SKIP) maps to four user intents: "buy now no thinking", "reasonable, deeper drop unlikely worth waiting", "hold, sale event probably coming", "don't buy at this price under any circumstance".
Where the 3% rule breaks down
Three caveats in the methodology page: brand-new releases (less than 90 days post-launch) don't have enough history for the threshold to be reliable. Titles with fewer than three sale events ever — common for niche indies — tend to behave erratically. And during major sale weeks (Summer Sale week one) the threshold can flip from STRONG_BUY to GOOD_BUY mid-day as the price moves.
ThereAny Score handles the first two by suppressing STRONG_BUY when price history is sparse. The third is unavoidable — recompute every 30 minutes is as fast as we can refresh without DDoSing CheapShark.
Frequently asked
- Why not use a percentile rank instead of an absolute threshold?
- Percentile rank breaks for titles with sparse price history (the 5th percentile of three observations is meaningless). Absolute distance from ATL is robust regardless of sample size and is what most shoppers reason about anyway.
- Does the 3% include or exclude tax / regional pricing?
- Excludes — we use base USD price as reported by CheapShark/ITAD. Regional pricing differences are bigger than the 3% band and would noise out the signal entirely.
- Is 30 days the right look-ahead window?
- We tested 7, 14, 30, 60, 90. Thirty days is the shortest window where the threshold-vs-followthrough signal is statistically clean. Shorter windows had too much weekend / Daily Deal noise; longer windows blurred sale-cycle structure.
- Will the threshold change?
- If the data tells us. We re-run the backtest annually. So far the 3% breakpoint has been stable across both Steam-driven sale years and post-pandemic catalogue-bloat years.