pv 衍生数据

pv数据

0阶数据 (原始)	一阶数据 (派生)	二阶数据 (高阶特征 & 技术指标)	三阶数据 (组合/因子)
open		cumret:`prod(1+returns)-1` excess_ret:`returns - rf` fear:`ts_mean(abs(returns - market_return)/(abs(returns)+abs(market_return)), 20)`	liquidity-adjusted return:`returns / (1+Illiquidity)`
close	returns:`(close - close[-1]) / close[-1]` — vwap:`sum(close*volume)/sum(volume)`	volatility:`ts_std(returns, 20)` RSI (relative strength index) MACD (EMA_diff(close)) ROC:`(close_t/close_{t-n} - 1)`	risk-adjusted momentum:`momentum / volatility`
high		maxdrawdown:`(peak(close) - trough(close)) / peak(close)` Williams %R:`(HH_n - close)/(HH_n-LL_n)`	volume-adjusted maxdrawdown:`maxdrawdown / adv20`
low		downside_vol:`ts_std(min(returns,0),20)` Stochastic %K/%D:`(close - LL_n)/(HH_n-LL_n)`	reversal-risk:`(-returns(1m)) / downside_vol`
volume	adv20:`ts_mean(volume, 20)` log_volume:`log(volume)` volume_z:`(volume - mean(volume,20))/std`	Illiquidity:`abs(returns) / volume` OBV (on-balance volume) CMF (Chaikin money flow)	liquidity_momentum:`momentum * (-Illiquidity)`
sharesout	mkt_cap:`close * sharesout` turnover:`volume / sharesout` turn20:`ts_mean(turnover, 20)`	size_effect:`log(mkt_cap)`	scaled_turnover:`turnover / size_effect`
vwap		spread:`(high - low)/vwap` Amihud illiq:`ts_mean(abs(returns)/dollar_volume,20)`	liquidity_volatility:`volatility * Amihud_illiq`
dollar_vol	dollar_volume:`close * volume`	turnover_vol:`ts_std(turnover,20)`	volume_volatility_ratio:`volatility / turnover_vol`
ohlc	hl_range:`(high - low)/close` oc_gap/overnight return:`(open - close[-1])/close[-1]`	realized_vol:`sqrt(sum(log(high/low)^2 + log(close/open)^2 ... over window))`	gap_volatility:`oc_gap / realized_vol`
returns		momentum:`ts_mean(returns, 12m)` reversal:`-returns(1m)`	momentum_reversal:`momentum * reversal`vol_mom:`momentum * (1/volatility)`
dividend 分红 split 拆股

rsv(未成熟随机值) = (close - ts_min(low, n)) / (ts_max(high, n) - ts_min(low, n)) n=9

对应关系

cap = close x sharesout
$\text{adj close} = \frac{close}{\prod\text{adj factor}}\approx \frac{close}{\sum(\text{adj factor}-1)+1}$
- adj_close 是为了表示的是close的一个长期的趋势, 连续的用adj_factor 进行调整
- 在相邻两天局部看, $\frac{\text{adj close}}{\text{adj close[-1]}}=\frac{close}{close[-1]}\cdot \frac{1}{\text{adj factor}}$
- $\prod\text{adj factor}=\prod(1+(\text{adj factor}-1))$ 只要一阶项就是后面那个加法的形式
adj_factor = 1/split(拆股日) , adj_factor = 1 - dividend/close_pre(除权日)
- adj_factor没有事件的时候是1, 有事的时候<1
- 因为我们想让不同时间的adj_close可比较
returns = (close/adj_factor-close_raw[-1])/close_raw[-1]
- 因为拆股和除权都是比较稀疏的事件,
Yahoo/Tiingo的adj close用的是后向复权(以最新日为锚点): 最新的adj close和实际的close更接近, 很久之前的adj close会很小

pv 衍生数据

Categories

Tags

Table of Contents

Related Posts

pv数据