clv-journal · git:20261031.825de9e · 2026-10-31 · sha256 ae970a915e943140
clv-journal git:20261031.825de9eA
Immutable. This exact content never changes and is served at /api/v1/blob/ae970a915e943140.
---
name: clv-journal
description: "Log bets with entry + closing odds, compute closing-line value (CLV) in odds pts and implied-prob pts, and roll up beat-close rate by sport/book/market."
homepage: https://github.com/ianalloway/openclaw-skills
metadata:
{
"openclaw": {
"emoji": "📈",
"requires": {
"bins": ["python3"]
}
}
}
---
# CLV Journal
Closing Line Value (CLV) is the process metric that predicts long-term betting edge. This skill keeps a **dedicated CLV journal** — entry price vs. close — separate from win/loss P&L. Beat the close consistently and the results follow.
Use alongside `bet-journal` (outcome/ROI) and `sports-odds` (live lines). This skill answers: *Did I get a better price than the market at close?*
## Formulas
**American → implied probability (no-vig single side):**
```
p = 100 / (a + 100) if a > 0
p = |a| / (|a| + 100) if a < 0
```
**Implied-prob CLV (pts):**
```
clv_prob = p_close − p_entry
```
Positive = you bought cheaper than the close (you beat the close). Example: entry −110 → p≈0.5238, close −120 → p≈0.5455 → CLV ≈ **+2.17 pts**.
**American odds CLV (pts):** for favorites (negative American), more negative close is better for the bettor who bought earlier:
```
clv_american_pts = entry − close # e.g. −110 vs −120 → +10 pts
```
For underdogs (positive American), higher entry than close means you beat the close:
```
clv_american_pts = entry − close # e.g. +150 vs +130 → +20 pts
```
**Spread / total line CLV (half-points):** if you bet a side, positive means the line moved in your favor after you bet:
```
# betting favorite −3.5 that closes −5.5 → +2.0 line pts
# betting underdog +3.5 that closes +1.5 → +2.0 line pts (same direction)
clv_line = signed_move_toward_your_side
```
**Beat-close rate:**
```
beat_close_rate = count(clv_prob > 0) / count(rows with both entry + close)
```
| Metric | Neutral | Good | Elite |
|--------|---------|------|-------|
| Avg CLV (prob pts) | ~0 | +1.0+ | +2.5+ |
| Beat-close rate | 50% | 55%+ | 60%+ |
| Sample size | — | 50+ | 200+ |
## Initialize the Journal
```bash
python3 -c "
import csv, os
JOURNAL = os.path.expanduser('~/.openclaw/clv-journal.csv')
os.makedirs(os.path.dirname(JOURNAL), exist_ok=True)
HEADERS = [
'id', 'ts_entry', 'ts_close', 'sport', 'event', 'market', 'side',
'book', 'odds_entry', 'odds_close', 'line_entry', 'line_close',
'clv_prob', 'clv_american', 'clv_line', 'beat_close', 'notes',
]
if not os.path.exists(JOURNAL):
with open(JOURNAL, 'w', newline='') as f:
csv.DictWriter(f, fieldnames=HEADERS).writeheader()
print(f'Created {JOURNAL}')
else:
print(f'Already exists: {JOURNAL}')
print('Markets: ML | SPREAD | TOTAL | PROP')
print('Leave odds_close / line_close blank until the market closes (pending).')
"
```
## Log an Entry (open price)
```bash
python3 -c "
import csv, os, uuid
from datetime import datetime, timezone
JOURNAL = os.path.expanduser('~/.openclaw/clv-journal.csv')
HEADERS = [
'id', 'ts_entry', 'ts_close', 'sport', 'event', 'market', 'side',
'book', 'odds_entry', 'odds_close', 'line_entry', 'line_close',
'clv_prob', 'clv_american', 'clv_line', 'beat_close', 'notes',
]
def log_entry(sport, event, market, side, book, odds_entry,
line_entry='', notes=''):
row = {
'id': str(uuid.uuid4())[:8],
'ts_entry': datetime.now(timezone.utc).strftime('%Y-%m-%dT%H:%M:%SZ'),
'ts_close': '',
'sport': sport,
'event': event,
'market': market,
'side': side,
'book': book,
'odds_entry': odds_entry,
'odds_close': '',
'line_entry': line_entry,
'line_close': '',
'clv_prob': '',
'clv_american': '',
'clv_line': '',
'beat_close': '',
'notes': notes,
}
with open(JOURNAL, 'a', newline='') as f:
csv.DictWriter(f, fieldnames=HEADERS).writerow(row)
print(f\"Logged entry {row['id']}: {sport} | {event} | {side} @ {odds_entry:+d} ({book})\")
print('Fill close later with log_close(id, odds_close, line_close=...)')
# --- EDIT THESE ---
log_entry(
sport='NBA',
event='Lakers @ Warriors',
market='SPREAD',
side='Warriors -3.5',
book='FanDuel',
odds_entry=-110,
line_entry=-3.5,
notes='early steam, sharp-leaning',
)
"
```
## Fill Closing Odds + Compute CLV
```bash
python3 -c "
import csv, os, tempfile
from datetime import datetime, timezone
JOURNAL = os.path.expanduser('~/.openclaw/clv-journal.csv')
def american_to_prob(a):
a = int(a)
return 100 / (a + 100) if a > 0 else abs(a) / (abs(a) + 100)
def clv_line_pts(side, line_entry, line_close):
if line_entry == '' or line_close == '':
return ''
le, lc = float(line_entry), float(line_close)
# Convention: side string starts with team; use signed lines as stored.
# Positive clv_line = line moved against the bettor's price (got better number).
# For a negative line (favorite), more negative close → positive CLV.
# For a positive line (dog), less positive / more negative close → positive CLV.
return round(le - lc, 2)
def log_close(bet_id, odds_close, line_close=''):
rows = []
found = False
with open(JOURNAL) as f:
reader = csv.DictReader(f)
fieldnames = reader.fieldnames
for row in reader:
if row['id'] == bet_id:
found = True
entry = int(row['odds_entry'])
close = int(odds_close)
p_e = american_to_prob(entry)
p_c = american_to_prob(close)
clv_p = round((p_c - p_e) * 100, 3) # probability points
clv_a = entry - close # American pts
clv_l = clv_line_pts(row['side'], row['line_entry'],
line_close if line_close != '' else row['line_close'])
row['ts_close'] = datetime.now(timezone.utc).strftime('%Y-%m-%dT%H:%M:%SZ')
row['odds_close'] = close
if line_close != '':
row['line_close'] = line_close
row['clv_prob'] = clv_p
row['clv_american'] = clv_a
row['clv_line'] = clv_l
row['beat_close'] = 'Y' if clv_p > 0 else ('T' if clv_p == 0 else 'N')
print(f\"Closed {bet_id}: entry {entry:+d} → close {close:+d}\")
print(f\" CLV prob: {clv_p:+.3f} pts\")
print(f\" CLV American: {clv_a:+d} pts\")
if clv_l != '':
print(f\" CLV line: {clv_l:+.2f} half-pts\")
print(f\" Beat close: {row['beat_close']}\")
rows.append(row)
if not found:
print(f'No row with id={bet_id}')
return
fd, tmp = tempfile.mkstemp(text=True)
os.close(fd)
with open(tmp, 'w', newline='') as f:
w = csv.DictWriter(f, fieldnames=fieldnames)
w.writeheader()
w.writerows(rows)
os.replace(tmp, JOURNAL)
# --- EDIT THESE ---
log_close('REPLACE', odds_close=-120, line_close=-5.0)
"
```
## One-shot: Entry + Close Together
When you already know both prices (post-game backfill):
```bash
python3 -c "
import csv, os, uuid
from datetime import datetime, timezone
JOURNAL = os.path.expanduser('~/.openclaw/clv-journal.csv')
HEADERS = [
'id', 'ts_entry', 'ts_close', 'sport', 'event', 'market', 'side',
'book', 'odds_entry', 'odds_close', 'line_entry', 'line_close',
'clv_prob', 'clv_american', 'clv_line', 'beat_close', 'notes',
]
def american_to_prob(a):
a = int(a)
return 100 / (a + 100) if a > 0 else abs(a) / (abs(a) + 100)
def log_pair(sport, event, market, side, book, odds_entry, odds_close,
line_entry='', line_close='', notes=''):
p_e, p_c = american_to_prob(odds_entry), american_to_prob(odds_close)
clv_p = round((p_c - p_e) * 100, 3)
clv_a = int(odds_entry) - int(odds_close)
clv_l = ''
if line_entry != '' and line_close != '':
clv_l = round(float(line_entry) - float(line_close), 2)
now = datetime.now(timezone.utc).strftime('%Y-%m-%dT%H:%M:%SZ')
row = {
'id': str(uuid.uuid4())[:8],
'ts_entry': now, 'ts_close': now,
'sport': sport, 'event': event, 'market': market, 'side': side,
'book': book,
'odds_entry': odds_entry, 'odds_close': odds_close,
'line_entry': line_entry, 'line_close': line_close,
'clv_prob': clv_p, 'clv_american': clv_a, 'clv_line': clv_l,
'beat_close': 'Y' if clv_p > 0 else ('T' if clv_p == 0 else 'N'),
'notes': notes,
}
with open(JOURNAL, 'a', newline='') as f:
csv.DictWriter(f, fieldnames=HEADERS).writerow(row)
print(f\"{row['id']} | {side} | {odds_entry:+d}→{odds_close:+d} | CLV {clv_p:+.3f} pts | beat={row['beat_close']}\")
# --- EDIT THESE ---
log_pair(
sport='NFL', event='Chiefs @ Bills', market='ML',
side='Bills', book='DraftKings',
odds_entry=-130, odds_close=-145,
notes='backfill from closing board',
)
"
```
## Rollup: Beat-Close Rate Dashboard
```bash
python3 -c "
import csv, os
from collections import defaultdict
JOURNAL = os.path.expanduser('~/.openclaw/clv-journal.csv')
if not os.path.exists(JOURNAL):
print('No journal. Run init first.'); raise SystemExit(1)
closed = []
pending = 0
with open(JOURNAL) as f:
for row in csv.DictReader(f):
if row['odds_close'] in ('', None):
pending += 1
continue
row['clv_prob'] = float(row['clv_prob'])
closed.append(row)
if not closed:
print(f'No closed rows yet ({pending} pending).'); raise SystemExit(0)
beats = sum(1 for r in closed if r['beat_close'] == 'Y')
ties = sum(1 for r in closed if r['beat_close'] == 'T')
avg = sum(r['clv_prob'] for r in closed) / len(closed)
print('=== CLV Journal Rollup ===')
print(f'Closed: {len(closed)} (pending: {pending})')
print(f'Avg CLV: {avg:+.3f} prob pts')
print(f'Beat-close rate: {beats}/{len(closed)} ({beats/len(closed):.0%}) ties={ties}')
print()
def rollup(key):
buckets = defaultdict(list)
for r in closed:
buckets[r[key] or '?'].append(r)
print(f'--- By {key} ---')
for k, rows in sorted(buckets.items(), key=lambda kv: -len(kv[1])):
b = sum(1 for r in rows if r['beat_close'] == 'Y')
a = sum(r['clv_prob'] for r in rows) / len(rows)
print(f' {k:<12} n={len(rows):<4} beat={b/len(rows):.0%} avg_clv={a:+.3f}')
print()
rollup('sport')
rollup('book')
rollup('market')
print('Recent (last 8):')
for r in closed[-8:]:
print(f\" {r['id']} {r['event']:<28} {r['side']:<18} {r['clv_prob']:+.3f} {r['beat_close']}\")
if avg > 1.0 and beats / len(closed) >= 0.55:
print()
print('VERDICT: Positive process — you are beating the close.')
elif avg > -0.5:
print()
print('VERDICT: Near-neutral CLV — tighten timing and line shopping.')
else:
print()
print('VERDICT: Negative CLV — chase earlier numbers or better books.')
"
```
## Quick Single-Bet CLV (no journal)
```bash
python3 -c "
def clv(odds_entry, odds_close):
def p(a):
a = int(a)
return 100 / (a + 100) if a > 0 else abs(a) / (abs(a) + 100)
pe, pc = p(odds_entry), p(odds_close)
pts = (pc - pe) * 100
print(f'Entry {odds_entry:+d} (p={pe:.4f}) → Close {odds_close:+d} (p={pc:.4f})')
print(f'CLV: {pts:+.3f} prob pts | American delta: {odds_entry - odds_close:+d}')
print('Beat close' if pts > 0 else ('Push' if pts == 0 else 'Lost to close'))
clv(-110, -125)
"
```
## Workflow Tips
1. **Log entry immediately** when you bet — do not wait for the result.
2. **Fill close at tip-off / first pitch / kickoff** from the same book (or a consensus close if you track one).
3. **Judge process on CLV**, not on the W/L of any single ticket.
4. Pair with `sports-odds` to capture live prices and with `bet-journal` for stake/ROI.
5. Need ≥50 closed bets before trusting beat-close rate; noise dominates small samples.
## Author
Created by [Ian Alloway](https://github.com/ianalloway) — Data Scientist specializing in sports analytics and ML.
## License
MIT License