You might find that price follows a pattern of correcting to the 0.786 retracement level before making new highs, but then throws in a curve-ball - correcting to, let's say, 0.5. In the future, it may go back to the 0.786 pattern again. In other words, a 50% retracement after a series of 21.4% retracments doesn't necessarily mean that that pattern is broken. It might just be one instance of breaking away from the pattern, as to throw off the market.
It's the number that keeps showing up the most that we're really after, not the most recent correction level. We use stop-losses with what we determine are to be good risk/ reward profiles just because we know that a pattern doesn't imply that it MUST HAPPEN again in the future. We're only trying to get probabilities to line up favorably with our trades, here.
The following are the retracement levels that I got: 0.786, 0.886, 0.786, 0.786, 0.786, 0.886, 0.886, 0.886, 0.886, 0.886, 0.786, 0.618, 0.886, 0.618, 0.8 (came really close to hitting 0.786, but never really touched it), 0.886, 0.886, 0.886, 0.786.
It would have been helpful to know that nearly every correction pulled back to the 0.886 to 0.786 range during that up-trend stretch. Could have designed a trade system off of it that would filter out trade signals that occur above what would be a 0.886 retracement at the current price level considered (perhaps instead wait for a breakout from a correction that is shallower than that, which should mean that it's a strong up-trend anyway, so the odds should still be in your favor). From there you could devise a lower risk at the 0.886 to higher risk at the 0.786, since the 0.886 stop is more likely to be taken out (perhaps add more to the position at the breakout from a shallow correction, which shows strength of trend).
In either case, stops about 25% above the next lower fibs level (I use 0.114, 0.272, 0.382, 0.5, 0.618, 0.787, and 0.886) would have worked.
Then you might wait for confirmation, or a little more confirmation, from your other technical analysis tools if/ when price breaks down below these two levels and tests the 0.618 on down to the 0.5, 0.382, 0.272, 0.114, 0.
It should be obvious by now, but what I mean by 0.886 retracement is actually 1 - 0.886, or 0.114 (11.4%).
This makes it apparent just how strong SPY's trend was, holding between 0.786 and 0.886 for most of its corrections and only correcting 0.618 twice, for that WHOLE TIME!
Using a buy at 0.786 pullback from the major swing pivot low to current high would have only stopped us out twice in this whole run. Of course, we can fine-tune the entry to use smaller price swing fibs levels, price patterns, support/ resistance levels, etc, to make more logical entries than simply buying on exactly a 0.786 retracement and to set-up smaller, more reasonable, stops (risk).