On Thursday, the system submitted a limit order, got filled, and exited the position three seconds later. The P&L: negative $0.16. The lesson: considerably more expensive than that.
Here is what happened. The Jun 5 plan called a bullish setup with a limit entry at $749.69 and a stop at the same level. The gate checked conditions at 9:45am and passed. The order went in. SPY opened volatile. The fill came back at $749.57, twelve cents below the stop. The bracket order saw the position immediately in stop territory and exited at $749.58. Three seconds, start to finish. The system behaved exactly as designed. The design had a gap.
The Scoreboard
Two weeks of live paper trading. All dollar figures are against a $100,000 starting balance per experiment.
- 001 Mancini Protocol: +$42.71 | Account: $100,042.71 | Record: 1-1
- 002 Overnight Drift: -$177.32 | Account: $99,822.68 | Record: 0-2
- 003 Congressional Mirror: +$243.92 | Account: $100,243.92 | 3 open positions
- SPY buy-and-hold benchmark: -$2,260.00 (-2.26%) over the same period
SPY dropped 2.26% since the experiments started. All three experiments are outperforming the benchmark. That includes the two with negative returns.
What I Noticed
The gate earned its keep twice this week.
Jun 2 and Jun 3 were both gated out. The plan said bullish, SPY opened strong, and the gate blocked entry both days because the FBD level never dipped. Price trended higher all session both days. Neither entry would have filled anyway. Zero missed valid entries.
Week 1’s question was whether the gate was being overly conservative. Week 2’s answer: no. The protection mechanism is filtering correctly. On trending days, the FBD mechanic structurally cannot trigger because the entry requires a pullback to reclaimed support. Trending markets and pullback entries are incompatible setups. The gate does not need to know this explicitly. It observes price, checks the condition, and kills the order if price has resolved away from the level.
Jun 4 showed what the strategy looks like when conditions are met.
SPY dipped to support at 9:36am. Filled at $753.67. Rallied to $756.94 by 12:11pm. Target hit, position closed, $42.59 profit. Five sessions of patience, one triggered trade, textbook outcome. The FBD setup works. It just requires specific market structure to activate: consolidation, a defined support level, and a dip back to that level after a breakout. Trending conditions produce zero fills. Ranging conditions produce the setup.
The Jun 5 loss was a bug report, not a trade.
The stop-undercut scenario is an edge case the system did not handle. The gate passed because SPY was at a level where entry made sense. But volatile opens create fills that deviate from the submitted price. When the fill lands below the stop, the bracket is already triggered on arrival. The position never had a chance to develop.
The correct behavior is a pre-submit check: if the current bid or ask is below the stop level at order submission time, cancel the order and log it as gated out rather than submitting. The fill risk was observable before the order went in. The system just did not look at that data point.
This is the part of building automated systems that does not appear in the planning phase. You model the happy path. You build the logic that executes it correctly. And then reality presents a scenario your model did not include. The scenario is not rare or exotic. It is a normal market microstructure outcome on a volatile open. The gap was in the model’s assumptions, not the market’s behavior.
The system found its own flaw. That is what paper trading is for.
Congress is doing something interesting.
Experiment 003 holds three positions: AAPL (Rep. Ed Case, D-HI, entered Jun 1 at $307.07), ARLP (Rep. Virginia Foxx, R-NC, entered Jun 3 at $25.45), and one test position. The account is at $100,243.92 with SPY at $737.55, down from $754.60 at experiment start. Congress is positive while the broad market is negative.
The hypothesis for Experiment 003 is that congressional disclosures carry information even with a mandatory 45-day filing lag. It is too early to draw conclusions from three positions in two weeks. But the positions not tracking SPY downward is at least consistent with the hypothesis that individual stock selection, not market beta, is where any edge would show up.
What to Watch in Week 3
The stop-undercut fix is the priority engineering task. Before any new order submits, the system needs to check: is the current price already at or below stop? If yes, kill the order and log as conditional rejection. The three-second trade should not be repeatable.
On the strategy side: SPY has been bearish this week, which structurally creates conditions for FBD setups. If Adam’s plan issues a bullish FBD setup into a pullback next week, the entry mechanic has a better environment to trigger than it did during last week’s trending sessions.
Follow @casualabsurdity for daily updates. Morning plans, gate decisions, and the full scoreboard post every trading day.
This is a paper trading experiment. Not financial advice.