POLYMARKET · PREDICTION MARKET · ATLANTA DREAM VS. TORONTO TEMPO

Atlanta Dream vs. Toronto Tempo: O/U 171.5

YES · live
36.5¢
NO · live
63.5¢

▸ Advanced metrics · M2M bundle

polymarket · wnba-atl-tor-2026-06-14-total-171pt5 · fresh · feed 0s old
24h sparkline · 60 pts
realized vol (ann.)
772.79%
max drawdown
7.21%
sharpe
ulcer index
4.59%
RMS drawdown
pain index
3.16%
mean drawdown
mod. VaR 95%
0.00%
Cornish-Fisher
martin ratio
ret / ulcer
CDaR 95%
7.21%
cond. drawdown
gain/pain
0.83
Σgain / Σ|loss|
sterling
ret / CDaR
omega (θ=0)
0.83
upside/downside
roll spread
2.8 bps
implied (price-only)
bars used
134
store
spread
24h Δ
flow lean
carry
flat
signalNEUTRALconfidence 20%
Same bundle via M2M API: /api/m2m/pm-wnba-atl-tor-2026-06-14-total-171pt5/bundle · venue execution: polymarket
LIVEPOLL0SRCFRESH8ms--:--:-- UTC8NEXT8.0sUP0s--:--HIST0/30
▶ STREAMING·HYPERLIQUID·POLYMARKET·0 POLLS·SRC FRESH·UPTIME 0s·NEXT POLL 8.0s·CC0 OPEN DATA·HYPO.MARKETS·▶ STREAMING·HYPERLIQUID·POLYMARKET·0 POLLS·SRC FRESH·UPTIME 0s·NEXT POLL 8.0s·CC0 OPEN DATA·HYPO.MARKETS·
YES · live
36.5¢
NO · live
63.5¢
YES price · live 24h
n=9 · μ=0.4856 · σ=0.0495 · range [0.3600, 0.5400] · R²=0.188 FALLING -28.00%σ HIGH 10.20%LAST 0.36000.54000.49500.45000.40500.3600μ = 0.4856max 0.5400min 0.3600dataMA(2)OLS R²=0.19μ lineμ ± σ bandmaxminlive endpoint
9 ticks · last 36.00¢
YES / NO split · live
YES 36.5%NO 63.5%NO63.5%63.50¢ · odds 1/1.57
Σ 100.00% · fair
Σ-sides total = 100.00% (tight rounding)
H(p) entropy = 0.947 / 1.00 bits (95%) · high uncertainty
YES
36.5%36.5¢2.74× +0.00pp
NO
63.5%63.5¢1.57× +0.00pp
Σ 100.00% · arb gap 0.00pp
Per-tick activity · |Δp| in basis points · live
n=8 · Σ=2,500 · μ=312.5 · σ=629.5 · CV=2.01BURSTY · concentratedcumulative energy ↗ · 50% by h=804509001,3501,800μ = 3131,80050%h1h2h3h4h5h6h7h8#1 peak#2-3> μactivequietμ linecum energy
Σ 2500bp moved · peak 1800bp · n=8 ticks
Live numerics · pulse on poll
LIVE NUMERICS8 metrics·POLL 0
snapshot age
8ms
YES mid
36.50¢ (36.50%)
NO mid
63.50¢ (63.50%)
ΣΣ sides
100.00%
arb gap
0.000pp
$24h vol $
$343.1k
liquidity $
$11.8k
history points
9 ticks (live)

§1 · 24h price history (YES + NO tokens)

YES price · CLOB mid
n=9 · μ=0.4856 · σ=0.0495 · range [0.3600, 0.5400] · R²=0.188 FALLING -28.00%σ HIGH 10.20%LAST 0.36000.54000.49500.45000.40500.3600μ = 0.4856max 0.5400min 0.3600dataMA(2)OLS R²=0.19μ lineμ ± σ bandmaxmin
9 YES observations from clob.polymarket.com · last 36.00¢
NO price · CLOB mid
n=9 · μ=0.5150 · σ=0.0496 · range [0.4600, 0.6400] · R²=0.195 RISING +28.00%σ HIGH 9.62%LAST 0.64000.64000.59500.55000.50500.4600μ = 0.5150max 0.6400min 0.4600dataMA(2)OLS R²=0.20μ lineμ ± σ bandmaxmin
9 NO observations from clob.polymarket.com · last 64.00¢

§2 · Distribution of Δp

Histogram of hourly increments
n=8 · 10 bins · μ=-0.0184 · σ=0.0587 · skew=-1.92 (left-skewed) · kurt=2.47 (leptokurtic (fat tails))653201-16.83ppbin -16.83pp · n=1 · 16.7% peakbin -16.83pp · n=1 · 16.7% peak-14.48pp-12.13pp-9.78pp-7.42pp-5.07pp-2.72pp6-0.37ppbin -0.37pp · n=6 · 100.0% peakbin -0.37pp · n=6 · 100.0% peak1.98pp14.33ppbin 4.33pp · n=1 · 16.7% peakbin 4.33pp · n=1 · 16.7% peakμΔ < 0 · loss barsΔ ≈ 0 · flatΔ > 0 · gain barsN(μ,σ²) referenceμ line · ±σ band shaded
n=8
Q-Q plot · standardised Δp vs N(0,1)
n=8 · skew=-1.83 · kurt=2.30 · near 2 / mid 5 / far 1 · OLS slope=0.85 intercept=-0.00LEPTOKURTIC — FAT TAILSUPPER TAIL NORMALFAT LOWER TAIL-3σ-3σ-2σ-2σ-1σ-1σ+0σ+0σ+1σ+1σ+2σ+2σ+3σ+3σsample ↓marginal: sample bars + theoretical N(0,1) curve →theoretical Φ⁻¹(p) →↑ sample z-quantile|Δ| < 0.3σ · on the line|Δ| < 1σ · moderate|Δ| ≥ 1σ · outliery = x refOLS fit
reference line = identity (perfect normality). Heavy upper-right tail = fat positive tail.

§3 · Sample moments

Descriptive statistics · 5-number summary · shape diagnostics
SAMPLE MOMENTS · N=9STRONGLY LEFT-SKEWED (G₁=-1.65)
μ MEAN48.56¢95% CI: [45.32¢, 51.79¢]
σ STD DEV4.95ppσ² = 24.528 · CV = 10.20%
med MEDIAN49.50¢Q₁ 49.50¢ · Q₃ 50.00¢
FIVE-NUMBER SUMMARY · BOX PLOT
min 36.00¢Q₁ 49.50¢med 49.50¢Q₃ 50.00¢max 54.00¢μ
SKEWNESS · G₁-1.652left-skewed
−3−10+1+3
EXCESS KURTOSIS · G₂1.755leptokurtic · fat tails
−30+2+4+6
μ ↔ medianμ < med · left-tailed|μ−med| / σ = 0.19
σ × 1.349 ↔ IQRdiverges from normalratio = 13.36
range ↔ σconcentrated (range < 4σ)range / σ = 3.63
μ = mean YES probability · σ = standard deviation · 95% CI = μ ± 1.96·SE. Skew/kurt diagnose departure from normality.

§5 · Time-series structure

Regime & autocorrelation diagnostics
TIME-SERIES STRUCTUREREGIME: MEAN-REVERTING · ρ(1) -0.30 + ADF rejected
ρ(1) AUTOCORR-0.304within white-noise band
ρ(2) AUTOCORR+0.028lag-2 not significant
H · HURST EXPONENT0.500random-walk
OLS TREND · t-STAT-1.272fails 5% test
HURST EXPONENT [0, 1]
H = 0.500RANDOM-WALK
0
anti-persistent
0.45
mean-reverting
0.5
random walk
0.55
persistent
1
strongly trending
AUTOCORRELATION FUNCTION · ρ(k) for k=1..5
k=1-0.304k=2+0.028k=3-0.026k=4-0.045k=5-0.0190+1−1+0.710.71+ momentum (ρ > +0.71)− reversal (ρ < −0.71)noise (within band)±2/√n threshold
OLS TREND · t-STAT · [-5, +5]
−5 reject−1.960 retain H₀+1.96+5 reject
REGIME CLASSIFICATIONMEAN-REVERTING · ρ(1) -0.30 + ADF rejectedfrom Hurst + ρ(1) joint diagnosis
PREDICTABILITY · score 0.30moderate · 1-step ahead inferrable|ρ(1)| + 2·|H − 0.5| heuristic
TREND SIGNIFICANCENOT SIGNIFICANT (|t|=1.27)α=0.05 critical |t|=1.96 · α=0.01 |t|=2.58
ρ(k) = lag-k sample autocorrelation · H = R/S Hurst exponent · t = OLS-trend t-statistic. Significance bands at ±2/√n approximate the 95% white-noise envelope. α=0.05 critical |t|=1.96; α=0.01 |t|=2.58.

§6 · Microstructure

Market quality · two-sided pricing · activity
MICROSTRUCTURE · MARKET QUALITYPERFECT · ARB-FREE Σ=100.00%
MARKET ID2542952
SLUGwnba-atl-tor-2026-06-14-total-171pt5
CATEGORYAtlanta Dream vs. Toronto Tempo
TWO-SIDED PRICING
PRIMARY · YES36.50¢implied prob 36.50% · decimal odds 2.74×
COUNTER · NO63.50¢implied prob 63.50% · decimal odds 1.57×
36.50¢
63.50¢
Σ-SIDES ARBITRAGE TEST
0%50%100% · target110%
Σ = 100.00% · |1 − Σ| = 0.000pp
24H ACTIVITY · LIQUIDITY
24H VOLUME343.10k USD 24h
LIQUIDITY11.78k USD
MARKET QUALITYPERFECT · ARB-FREE Σ=100.00%|1−Σ| ≤ 0.5pp ⇒ fair · > 2pp ⇒ inefficient
PRICING SKEWFAVOURS NO (64¢)|primary − counter| = 0.270 · entropy 0.947 bits
LIQUIDITY DEPTHDEEP100k+ deep · 10k+ active · 1k+ modest · 100+ thin
Σ-sides = YES + NO implied probabilities. Perfect arb-free Σ = 100%. |1−Σ| > 2pp suggests synthetic outright arbitrage.

§7 · Position sizing & edge analysis

Probability split · YES vs NO · Kelly · entropy · arbitrage
FAIR MARKET · no edge
YES 36.5%NO 63.5%YES36.5%H = 0.947 / 1.00 bits
Probability scale (YES)
0%25%50%
fair
75%100%
Implied decimal odds
YES2.74×(37¢)NO1.57×(64¢)
Kelly bet-size (% of bankroll) K* = 0.00%
K* full
0.00%
½K half
0.00%
¼K quarter
0.00%
Entropy H(p̂) = 0.947 bits (95% of max) · high uncertainty
0 (certain)0.250.50.751.00 (max)
Σ-sides = 100.00% · |1 − Σ| = 0.00pp · tight cross-venue rounding
K* full = (b·p − q)/b · ½K and ¼K are conservative fractions of the full-Kelly bet. Entropy in bits — log₂(2)=1 is maximum uncertainty for a binary market.

§9 · Hourly return heatmap

24-hour signed Δp grid · green = up · red = down
HOURLY RETURN HEATMAP · n=8 bars · best 5.50% · worst -18.00% · typical |Δ| 3.13%BEARISH SESSION -14.00%BEST+5.50%7hWORST-18.00%8hTYPICAL |Δ|3.13%mean absoluteCUMULATIVE-14.00%Σ signed ΔSTREAK↘ 1down-runASIA · 00-08 UTCμ +0.57% · Σ +4.00%EUROPE · 08-16 UTCμ -18.00% · Σ -18.00%US · 16-24 UTCμ n/a · Σ +0.00%CUMULATIVE Δ PATH · final -14.00%+4.00%-14.00%0.00% · 1h0.00% · 1h·1h0.00% · 2h0.00% · 2h·2h-0.50% · 3h-0.50% · 3h-0.50%3h0.00% · 4h0.00% · 4h·4h0.00% · 5h0.00% · 5h·5h-1.00% · 6h-1.00% · 6h-1.00%6h5.50% · 7h5.50% · 7h5.50%7h★ BEST-18.00% · 8h-18.00% · 8h-18.00%8h▼ WORSTTIME PATTERNAsia-led (+4.00%)RUNSup max 1 · down max 1BREADTH13% up · 38% down · 50% flat
1 up bars · 3 down · best 5.50% · worst -18.00% · typical |Δ| 3.125%

§10 · Equity curve & underwater drawdown

Cumulative compounded return + running peak-to-trough
EQUITY & DRAWDOWN ANALYSIS · n=9 barsSEVERE DRAWDOWN -14.78%FINAL-14.78%MAX DD-18.00%RECOVERYONGOING · 1 barsMAX RUN-UP+3.92%UNDERWATER5/9 (56%)STREAK↘ 1EQUITY CURVE · end 0.8522 · peak 1.0392 · range [0.8522, 1.0392]1.03920.8522break-even = 1★ PEAK 1.0392UNDERWATER DRAWDOWN · max -18.00% · severe0%-18.00%▼ TROUGH -18.00%TOP DRAWDOWN PERIODS · 2 total#1 -18.00%bar 9-9 · 1 bars · ONGOING#2 -1.50%bar 4-7 · 4 bars · recoveredDD SEVERITYsevere (max -18.00%)RECOVERYongoing · 1 barsTIME UNDER WATER56% of session · 5/9 bars
final equity 0.8522 (-14.78%) · max DD -18.00% · time-under-water 5/9 bars

§11 · Rolling-window statistics (w = 4 bars)

Rolling annualised Sharpe ratio · green positive · red negative
n=5 · +1 / −4 (20% positive) · μ=-32.47 · σ=40.99UNPROFITABLE STRATEGYLAST -31.09 (+0.03σ vs μ)73.3236.660.00-36.66-73.32μ = -32.47-46.80-46.80-46.80-46.80-73.32-73.3235.6435.64-31.09-31.09v > 0 · positivev < 0 · negativeμ mean lineμ ± σ bandlatest bar (outlined)
latest -31.090 · range [-73.32, 35.64] · μ -32.473 · positive Sharpe = excess-return-per-risk earned by buying-and-holding through this window
Rolling annualised volatility (%)
n=5 · μ=263.8141 · σ=398.7067 · range [23.3987, 950.9403] · R²=0.699 RISING +3964.07%σ EXTREME 151.13%LAST 950.9403950.9403719.0549487.1695255.284123.3987μ = 263.8141max 950.9403min 23.3987dataMA(2)OLS R²=0.70μ lineμ ± σ bandmaxmin
latest 950.94% · range [23.40%, 950.94%] · μ 263.81% · σ̂ scaled to annualised (×√8760)
Rolling lag-1 autocorrelation ρ(1)
n=5 · +0 / −5 (0% positive) · μ=-0.316 · σ=0.104MEAN-REVERSIONLAST -0.325 (-0.09σ vs μ)0.4170.2080.000-0.208-0.417μ = -0.316-0.417-0.417-0.417-0.417-0.205-0.205-0.215-0.215-0.325-0.325v > 0 · positivev < 0 · negativeμ mean lineμ ± σ bandlatest bar (outlined)
latest -0.325 · |ρ| > 0.3 ⇒ regime with persistence (ρ > 0) or reversal (ρ < 0) · |ρ| ≤ 0.1 = consistent with random walk

§12 · Hypothesis tests (α = 0.05)

Formal inference at 5% significance
1 of 4 REJECT · mixed evidence1 reject·3 pass·2 n/a·α = 0.05
𝒩

Jarque-Bera

REJECT H₀***

H₀: Δp ~ Normal(μ, σ²)

STATISTIC
19.8143
p-VALUE (log scale)
< 0.0001
α
10⁻⁴10⁻³10⁻²10⁻¹1
p < α · rejection zonenon-normal · fat tails or skew present
ρ

Ljung-Box(h=5)

FAIL TO REJECTns

H₀: No serial autocorrelation up to lag 5

STATISTIC
1.0653
p-VALUE (log scale)
0.5928
α
10⁻⁴10⁻³10⁻²10⁻¹1
p ≥ α · null retainedconsistent with white noise
Ψ

Dickey-Fuller (τ_μ)

N/An/a

H₀: p has a unit root (non-stationary)

STATISTIC
p-VALUE (log scale)
no decision possibleinsufficient data
±

Wald-Wolfowitz runs

N/An/a

H₀: Sign sequence of Δ is random

STATISTIC
p-VALUE (log scale)
no decision possibleinsufficient sign variety (1+/3-)
χ

KPSS (μ stationarity)

FAIL TO REJECTns

H₀: p IS level-stationary

STATISTIC
0.0000
p-VALUE (log scale)
1.0000
α
10⁻⁴10⁻³10⁻²10⁻¹1
p ≥ α · null retainedstationary not rejected (crit 0.463)
χ

Variance ratio q=1

FAIL TO REJECTns

H₀: Δp is a random walk · VR = 1

STATISTIC
0.0000
p-VALUE (log scale)
1.0000
α
10⁻⁴10⁻³10⁻²10⁻¹1
p ≥ α · null retainedVR 1.000 ≈ 1 (RW behaviour)
Each row states an explicit null H₀, the test statistic, an approximated p-value, and the decision. REJECT means evidence against H₀. KPSS complements ADF (rejecting both ⇒ ambiguous; rejecting one ⇒ clean verdict).

§13 · Spectral analysis (DFT periodogram)

Power spectrum of Δp · ‖X̂(k)‖²/n
n=4 bins · noise floor μ=5.04e-3 · top T=2.00h (35.7%) · top-3 cover 87.6%WHITE NOISE · no dominant cyclecumulative energy ↗ (0 bins above 2× noise)7.2e-35.4e-33.6e-31.8e-30.0e+0μ noise floorperiod 8.0 · power 2.50e-3 · 12.4% energyperiod 8.0 · power 2.50e-3 · 12.4% energyperiod 4.0 · power 3.93e-3 · 19.5% energyperiod 4.0 · power 3.93e-3 · 19.5% energyperiod 2.7 · power 6.53e-3 · 32.4% energyperiod 2.7 · power 6.53e-3 · 32.4% energyperiod 2.0 · power 7.20e-3 · 35.7% energyperiod 2.0 · power 7.20e-3 · 35.7% energy50% by T=2.7h#1 dominantT=2.00h#2T=2.67h#3T=4.00hT=2hT=3hT=4hT=6hT=8h← shorter cycle (high freq · Nyquist=½) · period T (bars per cycle) · longer cycle (low freq · 1/n) →#1 dominant#2 peak#3 peak> 2× noisenoiseμ floor2μ sig.cum energy
dominant period ≈ 2.00h (freq 0.500) · concentrates 35.7% of total energy · Σ|X̂|²/n = 2.015e-2

▸ Depth section using sovereign-store price series (134 bars · effective 1753297 bars/year) — annualisation reflects native polling cadence, not upstream timeframes.

§14 · Honest position analytics

A binary-market analytics module framed in horizon time (days to resolution, not annualised). Estimators that need a model probability q as a first-class input (Kelly, KL divergence, Bayesian posterior, Mark-to-Market MC) only render when q is provided externally. Sweep an exploratory q at the interactive simulator →

§15 · Horizon returns

Returns · per bar / per day / per horizon
Horizon 0.3 d · σ/bar 0.584pp · expected |Δp| over horizon 1.43ppterminal variance p(1−p) = 0.2480 · n = 134n = 134
μ per bar
-0.008pp
average Δp · drift
σ per bar
0.584pp
one-bar volatility · logit-free
Per-day movedaily
2.86pp
σ × √24
Per-horizon move0d
1.43pp
σ × √6
Terminal variancebinary
0.2480
p(1−p) at resolution
Current pricep
54.5¢
latest snapshot
Note: annualised Sharpe/Sortino are omitted — they are not meaningful for a bounded fixed-horizon binary contract that snaps to {0, 1} at resolution.
Annualised metrics are intentionally omitted — they don't apply to bounded probability series that resolve at a fixed date.

§16 · Tail risk

VaR · ES · max drawdown
VaR₉₅ 0.97pp · ES₉₅ 1.21pp · method parametric · drift-correcteddrift -0.008pp/bar · quantised: yes · median step 1.00pp · unique ratio 0.03low confidence · n < 200
VaR 95%
0.97pp
1.645·σ (parametric) of Δp
ES 95%
1.21pp
mean of the tail
Max drawdown
7.2pp
peak 55.5¢ → trough 51.5¢
Median step
1.00pp
price bucket granularity
Price series is bucketed (cent grid). Empirical quantiles collapse to grid points — parametric N(0, σ²) used instead.
Empirical quantiles unless the price series is bucketed (PM cent grid), in which case parametric N(0, σ²) is used to avoid grid collapse.

§17 · Odds conversion

Odds conversion · every dialect a bettor thinks in
Implied probabilityP
36.5%
= price
Decimal oddsEU
2.740
total return per $1
AmericanUS
+174
$100 wins $174
FractionalUK
1.74 / 1
profit per $1 risked
Profit per $100stake
+$173.97
clean dollar framing
-1000-5000+500+1000020406080100you · 36.5%implied probability (%)American odds
underdog (+)favorite (-)your price
Price → implied probability → decimal odds → American moneyline → fractional. Five views of the same number, plus the moneyline curve.

§18 · Binary entropy

Binary entropy · uncertainty as bits of information
Market entropyH(p)
0.947 bit
max 1.0 at p = 0.5
Your entropyH(q)
0.947 bit
Δ +0.000 bit vs market
Surprise · YES−log₂ p
1.45 bit
self-information
Surprise · NO−log₂(1−p)
0.66 bit
self-information
0.000.260.530.791.050.00.20.40.60.81.0marketmodelprobabilityH (bits)
Market entropy only — model entropy requires an external q.

§19 · Model-dependent surfaces

§ Edge / Kelly / KL · no model probability provided

External model required

The position-economics, Kelly, KL-divergence, Bayesian and Monte-Carlo surfaces require a model probability q as input — a number independent of the market price p.

The previous build defaulted q to a tape-momentum heuristic derived from p; that produces apparent edge that is structurally guaranteed to be small and is not a useful skill signal. The auto-derived path has been removed.

To explore these surfaces with a hypothetical q, open the interactive simulator and drag the MODEL P(YES) slider. To wire a real model, POST to the NOSTRADAMUS hook (TBD) or pass ?q=… on the simulator URL.

§∞ · Provenance & attestation

Upstream (snapshot)
gamma-api.polymarket.com
Upstream (history)
clob.polymarket.com
YES token ID
77229916553122518525281865034283824393452472387380187966521739454631079871575
NO token ID
8314323459929645260051762542078982319792357313377426316096362412624685532438
Snapshot fetched
2026-06-14 20:28:20 UTC
Snapshot age
8ms
History points
9 CLOB mids
Page rendered
2026-06-14 20:28:20 UTC
Storage policy
no persistence — fetched on every request
SHA-256 attestation
77ecc9b2cd548d4510c3574906d17c67de3bd97941916997c5c54610a205bf94 · deterministic hash of source snapshot
Open data licence
CC0 / public domain

§∞-2 · Related markets · explore more

Also see: /arb opportunities · RSS feed · more in Atlanta Dream vs. Toronto Tempo

Market depth

live order book · Polymarket YES
Depth within 1bp
$0
bid $0 · ask $0
Depth within 5bp
$0
bid $0 · ask $0
Depth within 10bp
$0
bid $0 · ask $0
Depth within 50bp
$0
bid $0 · ask $0
Mid price
0.365000
(best bid + best ask) / 2
Spread
5753.4bp
(bestAsk − bestBid) / mid
Imbalance (whole book)
-0.383
ask-heavy
Imbalance (top-5)
+0.645
bid-heavy top-of-book

Slippage scenarios

live book walk · Polymarket YES

Simulating a market order at three notionals against the live book. Slippage = avg execution price vs. mid, in basis points. Worst fill = price of the deepest level touched. Live JSON: /api/asset/pm-wnba-atl-tor-2026-06-14-total-171pt5/slippage?size=10000&side=buy

SideNotionalAvg fillSlippageWorst fillLevelsStatus
BUY$1.00K0.5785005849.31bp0.5900005FILLED
BUY$10.00K0.75753610754.40bp0.95000016FILLED
BUY$100.00K0.95345416122.03bp0.99000020PARTIAL
SELL$1.00K0.0813597770.99bp0.0800005FILLED
SELL$10.00K0.0574548425.91bp0.01000012PARTIAL
SELL$100.00K0.0574548425.91bp0.01000012PARTIAL

Risk metrics

sovereign store · 134 barsperiods/year ≈ 1.75M
Realized vol (annualised)
1433.80%
σ per bar = 0.010828
Mean return (annualised)
-23969.18%
μ per bar = -0.000137
Sharpe (rf=0)
-16.72
annualised; risk-free assumed zero
Max drawdown
7.21%
peak 0.56 → trough 0.52 over 16 bars

/api/asset/pm-wnba-atl-tor-2026-06-14-total-171pt5/risk · same metrics, JSON