In one sentence: Phil closes the technical-indicators detour by explaining the stochastic oscillator, admits a monthly-setting backtest on SPY looked good, warns that hindsight flatters it, and says a value investor who buys individual companies doesn't need any of it.
Key ideas
- Why bother. Danielle says that if she held an index she'd fret over every dip. Phil's coin-flip image: a crude rule can break paralysis at a fork even when it isn't predictive. [00:00–02:00]
- The stochastic oscillator. A momentum indicator that tries to say whether a market is overbought or oversold from price history. A common setting is 14 and 5 periods, with lines at 80% and 20%. Phil points to Investopedia for the maths. [02:00–04:00]
- How the signal works. A downward cross of the 80 line is read as "overbought, get out"; an upward cross of the 20 line as "oversold, get in". It can sit above 80 for a long time. The logic: after a long climb the buyers have all bought. [05:00–07:30]
- Setting it up. Brokerage charting tools (Thinkorswim calls them studies) offer these. Phil sets them to monthly periods: weekly is too jumpy, three or six months too slow. He says monthly roughly matches the daily death cross. [07:30–11:00]
- His settings. The three tools together: a moving average, MACD (he tweaks 12-26-9 to 9-17-8) and the stochastic at 14-5. He stresses this is his opinion only. [10:00–12:00]
- The backtest claim. Phil says that since 1999, exiting SPY on red arrows and re-entering on green would have roughly tripled an account by 2018. He immediately calls it a hindsight result that may not hold going forward. Treat the "tripled" figure as his claim. [13:00–15:30]
- Where it fails. In a sideways market you get whipsawed: out just before it rises, in just before it falls, and you pay costs each way. [16:00–17:00]
- The conclusion. If you own individual companies, you decide from price versus value and the story, so these tools aren't needed. They are only for people who hold an index and have no other exit rule. [17:00–18:30]
How it maps to RuleOne
- RuleOne's exit logic is value-based: sell when price passes sticker price or the story breaks. See the holdings view on /holdings/ and the sticker-price check on /stock/TICKER/.
- The screen has no oscillator and no crossover signal on purpose.
- Backtesting needs care: the same hindsight issue applies to any screen you run on past data.
Buffett, Munger and Graham links
- Graham on the market's moods and why price isn't a forecast: The Intelligent Investor, chapter 8 (Mr. Market).
- Buffett has long said he pays no attention to technical charts; Phil says so here. I'm not citing a specific letter.
- Munger's inversion habit fits Phil's own caution: ask how a backtest could be wrong before you trust it.
Words to know
- Stochastic oscillator: a momentum indicator flagging overbought (above 80) and oversold (below 20) readings.
- Whipsaw: a signal that flips back and forth in a flat market, costing you in trades.
- Backtest: applying a rule to past data to see how it would have done.
- Hindsight bias: seeing patterns clearly only after the fact.
Try this
Take a stock you own on /stock/TICKER/ and write down, in two lines, what would make you sell based on price against sticker price and what would make you sell based on the story. Do not use a chart signal.
Check yourself
- What does a stochastic reading above 80 claim?
Answer
That the market looks overbought; a downward cross of 80 is read as a sell signal. - Why is a backtest not proof?
Answer
Patterns found in hindsight may not repeat, and trading costs and whipsaws in flat markets can wipe out the gain. - Why doesn't a value investor need these?
Answer
Buy and sell decisions come from price versus value and the business story, not price patterns.
Short quotes
"Set the periods to monthly and slow this whole thing down." (Phil, ~09:50, auto-transcribed)