File size: 2,654 Bytes
d4c51f2
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
# Copy this to config.yaml and edit. All values are safe defaults for the
# "research + backtest, <$1k, conservative" profile.

data_dir: data

discover:
  # How to source candidate wallets. Any combination of the three.
  use_leaderboard: true          # pull top traders from the leaderboard
  leaderboard_windows: [all, 30d] # time windows to pull
  leaderboard_top_n: 200
  use_market_holders: true       # collect big holders of active high-volume markets
  markets_to_scan: 60            # how many active markets to inspect for holders
  holders_per_market: 40
  seed_wallets: []               # optional: paste addresses you already trust

analyze:
  max_wallets: 400               # cap to keep runs cheap
  activity_page_limit: 500       # data-api page size
  max_pages_per_wallet: 20       # hard cap on history depth per wallet
  request_sleep_sec: 0.25        # politeness delay between calls

# --- Screening: "long-term consistent profit + repeatable pattern" ---
screen:
  min_total_pnl_usd: 500         # must have made real money
  min_roi: 0.10                  # >= 10% return on deployed capital
  min_win_rate: 0.52             # better than coin flip on resolved markets
  min_resolved_markets: 40       # enough sample size
  min_active_days: 90            # "long-term", not a lucky week
  max_drawdown_pct: 0.5          # reject blow-up-prone wallets
  min_monthly_profit_ratio: 0.55 # >=55% of active months green (consistency)
  min_repeatability: 0.5         # pattern-repeatability score in [0,1]
  max_rebate_share: 0.35         # reject wallets whose profit is >35% rebates (market makers = uncopyable)
  top_n_shortlist: 25

# --- Copy-trade backtest ---
backtest:
  our_bankroll_usd: 800
  fee_bps: 0                     # Polymarket has no maker/taker fee today; keep 0
  slippage_bps: 150             # assumed adverse fill vs target's price (1.5%)
  copy_delay_sec: 60             # we react N seconds after the target's fill
  # Conviction-based position sizing (倉位跟著把握程度走)
  sizing:
    max_frac_per_trade: 0.05     # never risk >5% of bankroll on one entry
    max_frac_per_market: 0.15    # aggregate cap per market
    min_frac_per_trade: 0.005    # floor so tiny-conviction trades still register
    # conviction blends how big the target bet (vs their bankroll) and their edge
    w_stake: 0.6                 # weight on target's stake-fraction signal
    w_edge: 0.4                  # weight on entry-price edge signal
    kelly_cap: 0.5               # fraction of full-Kelly we allow (safety)
    assumed_max_edge: 0.15       # true-vs-price gap we credit the target at full conviction