diff --git a/mpl/fibionacci2.py b/mpl/fibionacci2.py
index 27b27f7..ec7896d 100644
--- a/mpl/fibionacci2.py
+++ b/mpl/fibionacci2.py
@@ -169,28 +169,6 @@ class Lohi:
             return None
 
 
-class Acp:
-    """Always correct predictor"""
-    def __init__(self, lows, highs):
-        self.lows = lows
-        self.highs = highs
-
-    def __call__(self, tick):
-        """Always correct predictor.
-        
-        Requires previous run of DelayedAcp to determine lows and highs.
-        
-        @param tick: The value of the current tick.
-        @type tick: tuple(n, cdt, last)
-
-        @return: 1. Tick if new max min has been detected,
-            2. None otherwise.
-        """
-        n, cdt, last = tick
-        res = None
-        return res
-
-
 def find_lows_highs(xs, ys):
     dacp = DelayedAcp(10)
     for tick in zip(range(len(xs)), xs, ys):
@@ -222,13 +200,13 @@ class DelayedAcp:
         res = None
         # automatic initialisation
         if self.mm0 is None:
-            # initalise water mark
+            # initialise water mark
             self.mm0 = tick
             res = self.mm0
             self.lows = [(n, cdt, last - 1)]
             self.highs = [(n, cdt, last + 1)]
         else:
-            # initalise trend until price has changed
+            # initialise trend until price has changed
             if self.trend is None or self.trend == 0:
                 self.trend = cmp(last, self.mm0[2])
         # check for max
diff --git a/mpl/fibionacci3.py b/mpl/fibionacci3.py
new file mode 100644
index 0000000..f714e6e
--- /dev/null
+++ b/mpl/fibionacci3.py
@@ -0,0 +1,420 @@
+# Copyright (c) 2008 Andreas Balogh
+# See LICENSE for details.
+
+""" animated drawing of ticks
+
+    using self.canvas embedded in Tk application
+    
+    Fibionacci retracements of 61.8, 50.0, 38.2, 23.6 % of min and max
+"""
+
+# system imports
+
+import datetime
+import os
+import re
+import logging
+import warnings
+import math
+
+import Tkinter as Tk
+import numpy as np
+
+import matplotlib
+matplotlib.use('TkAgg')
+
+from matplotlib.backends.backend_tkagg import FigureCanvasTkAgg
+import matplotlib.pyplot as plt
+import matplotlib.dates as mdates
+from matplotlib.dates import date2num
+
+# local imports
+
+# constants
+
+ONE_MINUTE = 60. / 86400.
+LOW, NONE, HIGH = range(-1, 2)
+
+# globals
+
+LOG = logging.getLogger()
+
+logging.basicConfig(level=logging.DEBUG,
+                    format='%(asctime)s.%(msecs)03i %(levelname).4s %(process)d:%(thread)d %(message)s',
+                    datefmt='%H:%M:%S')
+
+MDF_REO = re.compile("(..):(..):(..)\.*(\d+)*")
+
+
+def tdl(tick_date):
+    """ returns a list of tick tuples (cdt, last) for specified day """
+    fiid = "846900"
+    year = tick_date.strftime("%Y")
+    yyyymmdd = tick_date.strftime("%Y%m%d")
+    filename = "%s.csv" % (fiid)
+    filepath = os.path.join("c:\\rttrd-prd-var\\consors-mdf\\data", year, yyyymmdd, filename)
+    x = [ ]
+    y = [ ]
+    fh = open(filepath, "r")
+    try:
+        prev_last = ""
+        for line in fh:
+            flds = line.split(",")
+            # determine file version
+            if flds[2] == "LAST":
+                last = float(flds[3])
+                vol = float(flds[4])
+            else:
+                last = float(flds[4])
+                vol = 0.0
+            # skip ticks with same last price
+            if prev_last == last:
+                continue
+            else:
+                prev_last = last 
+            # parse time
+            mobj = MDF_REO.match(flds[0])
+            if mobj is None:
+                raise ValueError("no match for [%s]" % (flds[0],))
+            (hh, mm, ss, ms) = mobj.groups()
+            if ms:
+                c_time = datetime.time(int(hh), int(mm), int(ss), int(ms) * 1000)
+            else:
+                c_time = datetime.time(int(hh), int(mm), int(ss))
+            cdt = datetime.datetime.combine(tick_date, c_time)
+            x.append(date2num(cdt))
+            y.append(last)
+    finally:
+        fh.close()
+    # throw away first line of file (close price from previous day)
+    del x[0] 
+    del y[0] 
+    return (x, y) 
+
+def num2sod(x):
+    frac, integ = math.modf(x)
+    return frac * 86400 
+
+class Lohi:
+    """Time series online low and high detector."""
+    def __init__(self, bias):
+        assert(bias > 0) 
+        self.bias = bias
+        self.low0 = None
+        self.high0 = None
+        self.prev_lohi = NONE
+        self.lohis = [ ]
+        self.lows = [ ]
+        self.highs = [ ]
+
+    def __call__(self, tick):
+        """Add extended tick to the max min parser.
+        
+        @param tick: The value of the current tick.
+        @type tick: tuple(cdt, last)
+
+        @return: 1. Tick if new max min has been detected,
+            2. None otherwise.
+        """
+        n, cdt, last = tick
+        res = None
+        # automatic initialisation
+        if self.low0 is None:
+            self.low0 = tick
+            self.lows.append((n, cdt, last - 1))
+        if self.high0 is None:
+            self.high0 = tick
+            self.highs.append((n, cdt, last + 1))
+        if last > self.high0[2]:
+            self.high0 = tick
+            if self.prev_lohi == NONE:
+                if self.high0[2] > self.low0[2] + self.bias:
+                    res = self.high0
+                    self.low0 = self.high0
+                    self.lows.append(self.high0)
+                    self.lohis.append(self.high0)
+                    self.prev_lohi = HIGH
+        if last < self.low0[2]:
+            self.low0 = tick
+            if self.prev_lohi == NONE:
+                if self.low0[2] < self.high0[2] - self.bias:
+                    res = self.low0
+                    self.high0 = self.low0  
+                    self.lows.append(self.low0)
+                    self.lohis.append(self.low0)
+                    self.prev_lohi = LOW
+        if self.high0[1] < cdt - ONE_MINUTE and \
+            ((self.prev_lohi == LOW and \
+              self.high0[2] > self.lows[-1][2] + self.bias) or
+             (self.prev_lohi == HIGH and \
+              self.high0[2] > self.highs[-1][2])):
+            res = self.high0
+            self.low0 = self.high0
+            self.highs.append(self.high0)
+            self.lohis.append(self.high0)
+            self.prev_lohi = HIGH
+        if self.low0[1] < cdt - ONE_MINUTE and \
+            ((self.prev_lohi == LOW and \
+              self.low0[2] < self.lows[-1][2]) or 
+             (self.prev_lohi == HIGH and \
+              self.low0[2] < self.highs[-1][2] - self.bias)):
+            res = self.low0
+            self.high0 = self.low0  
+            self.lows.append(self.low0)
+            self.lohis.append(self.low0)
+            self.prev_lohi = LOW
+        if res:
+            return (self.prev_lohi, res)
+        else:
+            return None
+
+
+def find_lows_highs(xs, ys):
+    dacp = DelayedAcp(10)
+    for tick in zip(range(len(xs)), xs, ys):
+        dacp(tick)
+    return dacp.lows, dacp.highs    
+
+
+class DelayedAcp:
+    """Time series max & min detector."""
+    def __init__(self, bias):
+        assert(bias > 0) 
+        self.bias = bias
+        self.trend = None
+        self.mm0 = None
+        self.lohis = [ ]
+        self.lows = [ ]
+        self.highs = [ ]
+
+    def __call__(self, tick):
+        """Add extended tick to the max min parser.
+        
+        @param tick: The value of the current tick.
+        @type tick: tuple(n, cdt, last)
+
+        @return: 1. Tick if new max min has been detected,
+            2. None otherwise.
+        """
+        n, cdt, last = tick
+        res = None
+        # automatic initialisation
+        if self.mm0 is None:
+            # initialise water mark
+            self.mm0 = tick
+            res = self.mm0
+            self.lows = [(n, cdt, last - 1)]
+            self.highs = [(n, cdt, last + 1)]
+        else:
+            # initialise trend until price has changed
+            if self.trend is None or self.trend == 0:
+                self.trend = cmp(last, self.mm0[2])
+        # check for max
+        if self.trend > 0:
+            if last > self.mm0[2]:
+                self.mm0 = tick
+            if last < self.mm0[2] - self.bias:
+                self.lohis.append(self.mm0)
+                self.highs.append(self.mm0)
+                res = self.mm0
+                # revert trend & water mark
+                self.mm0 = tick
+                self.trend = -1
+        # check for min
+        if self.trend < 0:
+            if last < self.mm0[2]:
+                self.mm0 = tick
+            if last > self.mm0[2] + self.bias:
+                self.lohis.append(self.mm0)
+                self.lows.append(self.mm0)
+                res = self.mm0
+                # revert trend & water mark
+                self.mm0 = tick
+                self.trend = +1
+        return (cmp(self.trend, 0), res)
+
+
+class Main:
+    def __init__(self):
+        warnings.simplefilter("default", np.RankWarning)
+        self.ylow = None
+        self.yhigh = None
+        self.advance_count = 1
+
+        self.root = Tk.Tk()
+        self.root.wm_title("Embedding in TK")
+        
+        # create plot
+        fig = plt.figure()
+        self.ax1 = fig.add_subplot(211) # ticks
+        self.ax2 = fig.add_subplot(212) # diff from polyfit
+        # ax3 = fig.add_subplot(313) # cash
+    
+        self.ax1.set_ylabel("ticks")
+        self.ax2.set_ylabel("polyfit diff")
+        # ax3.set_ylabel("cash")
+    
+        major_fmt = mdates.DateFormatter('%H:%M:%S')
+        self.ax1.xaxis.set_major_formatter(major_fmt)
+        self.ax1.format_xdata = major_fmt
+        self.ax1.format_ydata = lambda x: '%1.2f' % x
+        self.ax1.grid(True)
+
+        self.ax2.xaxis.set_major_formatter(major_fmt)
+        self.ax2.format_xdata = major_fmt
+        self.ax2.format_ydata = lambda x: '%1.2f' % x
+        self.ax2.grid(True)
+        
+        # rotates and right aligns the x labels, and moves the bottom of the
+        # axes up to make room for them
+        fig.autofmt_xdate()
+
+        # create artists
+        LOG.debug("Loading ticks...")
+        self.x, self.y = tdl(datetime.datetime(2009, 6, 3))
+        LOG.debug("Ticks loaded.")
+        lows, highs = find_lows_highs(self.x, self.y)
+        
+        self.mmh = Lohi(10)
+    
+        self.w0 = 0
+        self.wd = 1000
+        self.low_high_crs = 0
+        xr, yr = self.tick_window(self.w0, self.wd)
+        fit = np.average(yr)
+        
+        self.tl, = self.ax1.plot_date(xr, yr, '-')
+        self.fl, = self.ax1.plot_date(xr, (fit,) * len(xr), 'k--')
+        self.mh, = self.ax1.plot_date(xr, (yr[0],) * len(xr), 'g:')
+        self.ml, = self.ax1.plot_date(xr, (yr[0],) * len(xr), 'r:')
+        # Acp markers
+        self.him, = self.ax1.plot_date([x for n, x, y in lows], [y for n, x, y in lows], 'go')
+        self.lom, = self.ax1.plot_date([x for n, x, y in highs], [y for n, x, y in highs], 'ro')
+
+        self.dl, = self.ax2.plot_date(xr, (0,) * len(xr), '-')
+
+        self.set_axis(xr, yr)
+
+        # embed canvas in Tk 
+        self.canvas = FigureCanvasTkAgg(fig, master=self.root)
+        self.canvas.draw()
+        self.canvas.get_tk_widget().pack(side=Tk.TOP, fill=Tk.BOTH, expand=Tk.TRUE)
+        
+        # toolbar = NavigationToolbar2TkAgg( self.canvas, self.root )
+        # toolbar.update()
+        # self.canvas._tkself.canvas.pack(side=Tk.TOP, fill=Tk.BOTH, expand=1)
+        
+        fr1 = Tk.Frame(master=self.root)
+        bu1 = Tk.Button(master=fr1, text='Quit', command=self.root.quit)
+        bu2 = Tk.Button(master=fr1, text='Stop', command=self.stop)
+        bu3 = Tk.Button(master=fr1, text='Resume', command=self.resume)
+        bu4 = Tk.Button(master=fr1, text='1x', command=self.times_one)
+        bu5 = Tk.Button(master=fr1, text='5x', command=self.times_five)
+        bu6 = Tk.Button(master=fr1, text='10x', command=self.times_ten)
+        bu1.pack(side=Tk.RIGHT, padx=5, pady=5)
+        bu6.pack(side=Tk.RIGHT, padx=5, pady=5)
+        bu5.pack(side=Tk.RIGHT, padx=5, pady=5)
+        bu4.pack(side=Tk.RIGHT, padx=5, pady=5)
+        bu2.pack(side=Tk.RIGHT, padx=5, pady=5)
+        bu3.pack(side=Tk.RIGHT, padx=5, pady=5)
+        fr1.pack(side=Tk.BOTTOM)
+
+
+    def animate(self):
+        self.w0 += self.advance_count
+        # prepare timeline window
+        xr, yr = self.tick_window(self.w0, self.wd)
+        while self.low_high_crs < self.w0 + self.wd:
+            self.mark_low_high(self.low_high_crs)
+            self.low_high_crs += 1
+        # build polynomial fit
+        lohis = self.mmh.lohis
+        if len(lohis) >= 4:
+            n0, x0, y0 = lohis[-4]
+            n1, x1, y1 = lohis[-1]
+            x2 = xr[-1]
+            coefs = np.polyfit([num2sod(x) for n, x, y in lohis[-4:]], [y for n, x, y in lohis[-4:]], 1)
+            self.fl.set_data((x0, x2), [np.polyval(coefs, num2sod(x)) for x in (x0, x2)])
+            # width of trend channel
+            mx = 0
+            for n in range(n0, n1):
+                mx = max(mx, math.fabs(np.polyval(coefs, num2sod(self.x[n])) - self.y[n]))
+            a, b = coefs
+            self.mh.set_data((x0, x2), [np.polyval((a, b+mx), num2sod(x)) for x in (x0, x2)])
+            self.ml.set_data((x0, x2), [np.polyval((a, b-mx), num2sod(x)) for x in (x0, x2)])
+        # update tick line
+        self.tl.set_data(xr, yr)
+        # update axis
+        self.set_axis(xr, yr)
+        self.canvas.draw()
+        if self.w0 < len(self.x) - self.wd - 1:
+            self.after_id = self.root.after(10, self.animate)
+            
+    def set_axis(self, xr, yr, bias=50):
+        if self.ylow is None: 
+            self.ylow = yr[0] - bias / 2
+            self.yhigh = yr[0] + bias / 2
+        for y in yr:
+            if y < self.ylow:
+                self.ylow = y
+                self.yhigh = self.ylow + bias
+            if y > self.yhigh:
+                self.yhigh = y
+                self.ylow = self.yhigh - bias
+        self.ax1.axis([xr[0], xr[-1], self.ylow, self.yhigh])
+        self.ax2.axis([xr[0], xr[-1], -25, +25])
+    
+    def tick_window(self, w0, wd = 1000):
+        return (self.x[w0:w0 + wd], self.y[w0:w0 + wd])
+        
+    def mark_low_high(self, n):
+        x = self.x
+        y = self.y
+        rc = self.mmh((n, x[n], y[n]))
+        if rc:
+            lohi, tick = rc
+            nlh, xlh, ylh = tick
+            if lohi < 0:
+                # low
+                self.ax1.annotate('low',
+                            xy=(x[nlh], y[nlh]),
+                            xytext=(x[n], y[nlh]),
+                            arrowprops=dict(facecolor='red',
+                                              frac=0.3,
+                                              shrink=0.1))
+            elif lohi > 0:
+                # high
+                self.ax1.annotate('high',
+                            xy=(x[nlh], y[nlh]),
+                            xytext=(x[n], y[nlh]),
+                            arrowprops=dict(facecolor='green',
+                                              frac=0.3,
+                                              shrink=0.1))
+
+    def stop(self):
+        if self.after_id:
+            self.root.after_cancel(self.after_id)
+            self.after_id = None
+    
+    def resume(self):
+        if self.after_id is None:
+            self.after_id = self.root.after(10, self.animate)
+
+    def times_one(self):
+        self.advance_count = 1
+
+    def times_five(self):
+        self.advance_count = 5
+    
+    def times_ten(self):
+        self.advance_count = 10
+    
+    def run(self):
+        self.root.after(500, self.animate)
+        self.root.mainloop()
+        self.root.destroy()
+
+
+if __name__ == "__main__":
+    app = Main()
+    app.run()